| mason | Hey all. I'm moving some VMs over from CentOS, and I've got my bridge set up, but I'm struggling to figure out how to aim virt-manager at it. Is attaching to an existing bridge possible with virt-manager? | 01:07 |
|---|---|---|
| pmatulis | mason, you have a ubuntu server acting as a KVM host and you want to manage its VMs from a desktop? | 01:08 |
| mason | pmatulis: Well, from a combination of virt-manager and virsh. | 01:08 |
| sarnold | hey mason :) | 01:09 |
| mason | sarnold: o/ | 01:09 |
| mason | pmatulis: I don't like the three pages of command line I need to define new VMs with virsh. I like the wizardly approach. | 01:09 |
| mason | Ah, maybe it's an issue with virt-manager. Just noted https://bugzilla.redhat.com/show_bug.cgi?id=1355907 | 01:11 |
| ubottu | bugzilla.redhat.com bug 1355907 in netcf "Unable to use / create existing NetworkManager bridge" [Unspecified,New] | 01:11 |
| mason | (although I made my bridge manually, not with NM) | 01:11 |
| pmatulis | mason, ok, so your KVM host has a bridge, say br0? | 01:12 |
| mason | It is indeed br0. | 01:12 |
| mason | (FWIW, I've spent years with Xen, so KVM is still fairly new to me. Also, bridging still seems slightly alien to me.) | 01:13 |
| pmatulis | mason, so you are now trying to configure a connection to the KVM host from virt-manager right? | 01:13 |
| mason | No, that's easy - I can connect to it. I want to be able to define new virtual machines using the existing bridge. | 01:14 |
| mason | If from the Virtual Network tab I say to add a new one - call it foo - and I turn off address space definition, I come to a choice of isolated virtual network for forwarding to a physical network. This has a popdown that doesn't end up listing br0, although it lists the underlying ethernet interface. | 01:15 |
| mason | It seems unhappy with this, and complains that I haven't provided an IP address for network 'foo' | 01:16 |
| pmatulis | mason, what ubuntu release is running on the desktop? | 01:17 |
| mason | virt-manager seems willing to create new bridges for me, but it's not seeing the existing one | 01:17 |
| mason | pmatulis: This is all on a single Xenial box. | 01:17 |
| mason | I'm not unwilling to edit virt-manager or libvirt config, but it seems odd that I'm running into a wall here. | 01:18 |
| pmatulis | mason, you mean the kvm host is running a graphical environment? | 01:18 |
| pmatulis | (just one machine?) | 01:18 |
| mason | pmatulis: It is, yes. nVidia proprietary drivers even. Sometimes Steam runs alongside the virtual machines. | 01:19 |
| mason | Yeah, not a cluster. | 01:19 |
| mason | But even so, it seems like I should be able to use the existing bridge. | 01:19 |
| mason | I'll go compare the old config from the CentOS environment. | 01:19 |
| pmatulis | i have not needed to configure that stuff in a while. it automatically chooses the bridge. although i've never run it on kvm host itself before | 01:20 |
| mason | pmatulis: So, under CentOS, it offered to make a bridge for me, and it set it up itself, but now that I'm back on Ubuntu with the very comfortable /etc/network/interfaces and friends, I just made the bridge myself. I would have run into this same issue on CentOS I guess. | 01:20 |
| mason | FWIW, the bridge config is largely identical to what I used for a long time on Debian/Xen, and it seems to be working fine. | 01:21 |
| sarnold | hrm, when I hit the 'specify shared device name' i get a field for bridge name | 01:21 |
| sarnold | but no idea how to use it; does it work? :) | 01:22 |
| mason | (Backstory: It's my desktop, and I use it as a lab environment for work reproductions as well.) | 01:22 |
| mason | sarnold: It's a freeform field I believe. | 01:22 |
| mason | looking | 01:22 |
| mason | sarnold: Wait, where did you find that? | 01:22 |
| mason | If I could "specify shared device name" I think that's where I'd point it at my bridge, but I haven't found something with that wording as yet. | 01:24 |
| sarnold | mason: 'open' a vm, hit the 'i' icon, focus the 'nic:xx:xx:xx' entry in the sidebar on the left; then the 'Network source' dropdown box | 01:25 |
| mason | Oh, I don't have any VMs defined yet. Hrm. | 01:26 |
| sarnold | ohhhhh | 01:26 |
| mason | Trying to set up the virtual network in advance. | 01:27 |
| sarnold | I just used what I had rather than going through the wizard | 01:27 |
| mason | That said... In the CentOS /etc/libvirt, I do only see the bridge defined in actual VM .xml files. | 01:27 |
| sarnold | bah and the wizard gets real personal real quick "give us an install media" uhhhh let me just click! | 01:27 |
| mason | heh | 01:27 |
| mason | I'll make a VM and see if I can back my way into the right bridge. | 01:28 |
| sarnold | aha, step 5 of 5, there's a weeeeee tiny little triangle near the text Network Selection | 01:29 |
| mason | That's where the "specify" field is. Trying it. | 01:29 |
| sarnold | hit that little triangle and there's a dropdown, change that to "Specify shared device name", and that adds a new text field to the dialog box | 01:29 |
| mason | Yep! | 01:30 |
| mason | So, *somewhere* in the config there exists a way to specify that so I can pull my bridge off the menu, rather than typing in br0 | 01:30 |
| sarnold | I <3 that a simple text field is hidden behind two separate "this is too advanced for you" things | 01:30 |
| mason | hehe | 01:30 |
| mason | So, under Xen/Debian I *loved* xen-tools to set up most of my defaults. Is there such a thing for KVM/libvirt under Ubuntu? | 01:31 |
| sarnold | no idea what xen-tools does.. the server team put together a uvt-tool that tries to abstract over a bunch of libvirt things | 01:31 |
| mason | I'll look at it. | 01:32 |
| mason | xen-tools lets you set up a set of prefs, so you only had to specify deviations when creating a new vm | 01:32 |
| sarnold | but I lose track of things right about the time I find out that I've got to hand-edit xml in order to use zfs datasets for backing devices | 01:32 |
| mason | half a sec | 01:32 |
| mason | sarnold: Oh no no no! | 01:32 |
| mason | virt-manager makes that easy | 01:32 |
| sarnold | does it? | 01:32 |
| sarnold | maybe I can skip my libvirt NIH then | 01:32 |
| mason | say "yeah, I want storage, but I'll specify the device" | 01:32 |
| mason | Do you use Xen? | 01:33 |
| sarnold | no | 01:33 |
| mason | Or just straight KVM? | 01:33 |
| sarnold | yeah | 01:33 |
| mason | Anyway, you get a freeform text field, and you can say things like: /dev/zvol/zroot/vm/foo | 01:33 |
| mason | ...which is what I did for this test VM. | 01:33 |
| mason | Works fine. | 01:33 |
| sarnold | for some reason the libvirt based tools never seem to work real well for me, so I set out to write my own qemu wrapper becase How Hard COuld It Be? three months later and it still doesn't boot any machines. lol. | 01:33 |
| mason | Also, FWIW, your "specify" option you noted works fine. Spun up my VM, and I could pull an address from DHCP. | 01:34 |
| sarnold | \o/ | 01:34 |
| mason | sarnold and pmatulis: If you're *not* using virt-manager, is there a reasonable way to get a console on a new VM you're building right off? | 01:34 |
| mason | Part of my using it is ignorance of the options. | 01:35 |
| mason | So I configure stuff with virt-manager, but then randomly start/stop things with virsh or virt-manager, depending on what's in front of my at the moment. | 01:35 |
| mason | s/my/me/ | 01:35 |
| sarnold | mason: one of my coworkers on the security team wrote a wrapper around the libvirt wrappers :) so the idea is you'd use 'uvt new precise amd64 hostname-here' to build the machine with defaults from a config file, then 'uvt start hostname-here', 'uvt stop hostname-here', 'uvt view hostname-here', etc | 01:39 |
| mason | Oh, right, I was going to get some config snippets, speaking of wrappers. | 01:39 |
| sarnold | mason: but the tool requires so much security-team specific tooling that it's probably worth looking at the server team's 'uvt-tool' instead. (completely different uvt. oops.) | 01:39 |
| mason | heh | 01:39 |
| mason | So, xen-tools has a config that lets me specify a default volume group (LVM-centric) for VM disks, lets me specify debootstrap or rinse or various install methods, lets me set default sizes for memory, swap, disk, fs types, default networking types. | 01:40 |
| sarnold | sounds lovely :) | 01:41 |
| mason | And then when I want to create something, I don't have to say much. Example: | 01:41 |
| mason | xen-create-image --hostname=FOO --mac=DE:AD:BE:EF:00:** | 01:41 |
| mason | or if I want to override debootstrap and make it a CentOS box, I could say: | 01:41 |
| mason | xen-create-image --force --hostname=centos --mac=DE:AD:BE:EF:00:06 --install-method=rinse --dist=centos-6 | 01:41 |
| mason | I'd love something similar for KVM/libvirt, so I'll look at that uvt-tool stuff. | 01:42 |
| sarnold | uvt-tool may be too specific to ubuntu | 01:42 |
| mason | Whenever I see how people specifying a metric tonne of options on a command line for virt-install, it kind of horrifies me. | 01:42 |
| mason | Maybe. I tend to run Ubuntu, FreeBSD, CentOS, and varieties of RHEL on this. | 01:43 |
| mason | So, compare my xen-create-image, above, with the example they give for virt-install here: https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Virtualization_Host_Configuration_and_Guest_Installation_Guide/sect-Virtualization_Host_Configuration_and_Guest_Installation_Guide-Guest_Installation-Creating_guests_with_virt_install.html | 01:44 |
| compdoc | i let virt-manager do all the work for me | 01:44 |
| mason | compdoc: Yar. The trick is that if virt-manager didn't define your bridge, you have to invoke Mickey Mouse to specify an existing bridge, it seems. | 01:44 |
| compdoc | I create the bridges manually | 01:44 |
| mason | compdoc: Same. Is there a way to not have to type the bridge name into virt-manager, so as to make that bridge the default? | 01:45 |
| mason | That's what started all this. :P | 01:45 |
| compdoc | all bridges and interfaces appear as drop-down lists | 01:45 |
| compdoc | err, no. as a list | 01:45 |
| mason | compdoc: It doesn't here - I have to specify a shared device. If I type my bridge name into that field, it does the right thing. | 01:46 |
| mason | The difference is that on CentOS I could have a default network selected, so I could whip past that. Not a big deal, but it struck me as an odd difference. | 01:46 |
| mason | compdoc: FWIW, it seems to exist elsewhere too: https://bugzilla.redhat.com/show_bug.cgi?id=1355907 | 01:47 |
| ubottu | bugzilla.redhat.com bug 1355907 in netcf "Unable to use / create existing NetworkManager bridge" [Unspecified,New] | 01:47 |
| compdoc | oh. heh. network manager. I tend not to install that on Server | 01:47 |
| sarnold | mason: at least that example's easy enough to shove into a shell script | 01:47 |
| mason | compdoc: Same here. But I'm not using it - the problem is the same though. | 01:47 |
| mason | sarnold: That's true. And I could wrap it easily enough. I'm just... lazy... I guess. :P | 01:48 |
| mason | Plus, I like popping right into a console, which virt-manager makes easy. | 01:48 |
| mason | In any event, I think I have enough to get my VMs moved over now. =cheers= | 01:48 |
| mason | I need to move this channel to a better window. It's /window 41 now. | 01:49 |
| sarnold | it's /win 30 for me | 01:49 |
| sarnold | not ideal, since #debian-security is /win 38 and the idfference between the two is mighty small | 01:49 |
| mason | There, not it's /window 4. I made #zfs and #openzfs share a window. They're both mostly dead, so it'll work. | 01:50 |
| mason | s/not/now/ | 01:51 |
| sarnold | hehe | 01:52 |
| mason | So, there were a lot of niggling little changes needed. Some examples: s/pc-i440fx-rhel7.0.0/pc-i440fx-xenial/g s/Skylake-Client/Broadwell/g were the big things. | 03:06 |
| mason | That said, my VMs are happily moved. | 03:06 |
| cpaelzer | good morning | 05:31 |
| === danpawlik_absent is now known as dpawlik | ||
| === Guest95054 is now known as lordievader | ||
| === lordievader is now known as Guest95840 | ||
| === Guest95840 is now known as lordievader | ||
| helpImStuck | So.. I want to create a lxc router where the physical nic are bound to the container and removed from rest of the system. Is this possible? | 08:25 |
| ikonia | helpImStuck: it would be quiet hard as the kernel is what provides netfilter and the container would need to interact with the kernel to update the rules | 08:32 |
| ikonia | helpImStuck: it seems like a bad idea, when a VM would work better if you require it virtualized | 08:33 |
| ikonia | also securing a containers interface is quite tricky | 08:33 |
| helpImStuck | so what are the containers good for? I tried alpine and acf . it was cool. | 08:36 |
| ikonia | alpine...ha ha ha ha ha ha ha ha ha ha | 08:46 |
| ikonia | helpImStuck: self contained mass immuatable deployment | 08:47 |
| helpImStuck | And what does that even mean xD immutable . | 08:49 |
| helpImStuck | helpImStuckAndMyEnglishSuck should i name myself here | 08:51 |
| ikonia | helpImStuck: "throw away" | 08:51 |
| helpImStuck | and why not alpine? They've created a small distro with.. hm. didn't work so well. small things like it didn't save my keymap after boot. And it's been along for a long time. | 08:52 |
| helpImStuck | it's good in theory | 08:53 |
| ikonia | small != good | 08:54 |
| helpImStuck | I like lxd 2 because i can run arch and have all the packages from yaourt on an ubuntu base.. | 08:54 |
| ikonia | what ? | 08:54 |
| helpImStuck | Yes, small, not so complex | 08:54 |
| ikonia | complex...it's very complex | 08:54 |
| ikonia | they have applied some custom security patches | 08:55 |
| ikonia | they have made their own "odd" package manager format | 08:55 |
| helpImStuck | are unprivileged lxc container safer? | 08:55 |
| helpImStuck | s | 08:55 |
| ikonia | safer than what ? | 08:56 |
| helpImStuck | alpine on xen | 08:57 |
| ikonia | I can't / won't comment on other setups like that | 08:57 |
| ikonia | #ubuntu-server is for ubuntu server based support | 08:57 |
| helpImStuck | this is that channel | 08:57 |
| ikonia | how secure/good alpine is isn't as a container guest isn't really for this channel | 08:57 |
| helpImStuck | but you can use alpine in a lxc container.. so it's related | 08:58 |
| ikonia | no it's not | 08:58 |
| ikonia | you can use almost any OS in a container | 08:58 |
| ikonia | how good that OS is as a container isn't really ubuntus issue | 08:58 |
| helpImStuck | and then, for someone to get support if they use ubuntu server + lxc and let's say arch.. they have to look for help on the arch wiki instead of here? Even tho both os:es are involved | 08:59 |
| ikonia | helpImStuck: if they are having problem with the arch container, yes | 09:00 |
| ikonia | if they are having problems with the ubuntu host hosting the container, no | 09:00 |
| helpImStuck | oh.. now i get it.. LXD is not ubuntu specific | 09:01 |
| helpImStuck | lxd/lxc | 09:01 |
| helpImStuck | it can't be. i'm using ubunt userver as host for the containers | 09:04 |
| helpImStuck | -u | 09:04 |
| ArchaicLord | Morning all! Could use some help and mentoring please.. I had a ubuntu server which was installed on a usb. It then had a seperate raid array which i was using for meida sharing. I moved house and as a result the easist thing seemed to be to reinstall the ubuntu. which i have now done. I pluged the raid disks back in and to my suprise unutu seems to have picked it up. The issue I have is rec | 10:00 |
| fallentree | The issue you have is unfinished senten | 10:06 |
| ArchaicLord | Sorry, its hard to see a lot of text in a tiny box | 10:10 |
| ArchaicLord | So fdisk -l produces Disk /dev/md0: 3.7 TiB, 4000529252352 bytes, 7813533696 sectors Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 4096 bytes I/O size (minimum/optimal): 524288 bytes / 1048576 bytes | 10:11 |
| ArchaicLord | this is the raid I previoulsy set up | 10:11 |
| ArchaicLord | and its been picked up automaticly | 10:12 |
| ArchaicLord | lv display shows the two volumes I created previously. will having different host names cause an issue | 10:13 |
| fallentree | hostnames? no, hostnames have nothing to do with software raid or LVM | 10:13 |
| ArchaicLord | or do I just need to configure SAMBA ? | 10:14 |
| fallentree | ArchaicLord: what problem do you have that you wish to solve? | 10:15 |
| === Daviey_ is now known as Daviey | ||
| ArchaicLord | I am trying to reimpliment the old array. | 10:15 |
| fallentree | define reimplement. From what you say, Ubuntu picked it up automatically. | 10:15 |
| ArchaicLord | yes it has.. I can see it I can view it locally, so I guess I just need to set up samba config to complete it | 10:16 |
| fallentree | yeah, if you want to export it over smb/cifs protocols | 10:17 |
| ArchaicLord | i have a feeling I may have mapped the lvm to another place on the previous setup .. | 10:38 |
| ArchaicLord | I can cd to /dev/file-server/ inside here i have 2 folders but when I can't cd into them. when I do ls -l IU get folder -> ../dm-0 /-1 | 10:39 |
| ArchaicLord | wow my typing sucks. | 10:40 |
| fallentree | ArchaicLord: LVM is mapped under /dev/mapper/ | 10:42 |
| fallentree | probably something like /dev/mapper/vg0/... where vg0 is the volume group name | 10:42 |
| fallentree | ArchaicLord: oh also, you have to mount the lv somewhere, the stuff under /dev are (mapped) block devices | 10:46 |
| ArchaicLord | would that be like mounting to md0 | 10:47 |
| fallentree | ArchaicLord: not 'to' but 'of'. eg. mount /dev/mapper/lv-foo /mnt/someplace | 10:48 |
| fallentree | you mount the lv inside vg on a pv :) | 10:48 |
| fallentree | (to a directory) | 10:48 |
| fallentree | iirc lvscan will tell you what the logical volumes you have available | 10:49 |
| fallentree | their labels are available as block devices under /dev/mapper/ | 10:49 |
| fallentree | it's been a while since I LVM'd | 10:49 |
| ArchaicLord | LOL.. sorry fallentree.. My trouble is I previously set it up over a year ago. I was doing it for the first them and found it hard. | 10:49 |
| ArchaicLord | now i have a fagiue memory and can't rember exaclty how it was set up | 10:50 |
| ArchaicLord | lvscan gives me this | 10:51 |
| ArchaicLord | ACTIVE '/dev/file-server/plex-share' [1.50 TiB] inherit ACTIVE '/dev/file-server/server-backup' [1.50 TiB] inherit | 10:51 |
| fallentree | I suppose those are the mountable names | 10:52 |
| fallentree | eg. try `mkdir -p /mnt/plex-share && mount /dev/file-server/plex-share /mnt/plex-share` | 10:53 |
| fallentree | same for server-backup, then see if you can access files under /mnt/plex-share and /mnt/server-backup | 10:53 |
| ArchaicLord | ahh so i needed to mount the directory | 11:01 |
| ArchaicLord | mounting to /mnt/ allows me to cd into the share | 11:01 |
| ArchaicLord | now i just need to sort out permissions | 11:02 |
| ArchaicLord | thank you | 11:12 |
| ArchaicLord | fallentree thank you all sorted.. In windows I can now access the foleder and create new files | 11:13 |
| fallentree | nice. | 11:13 |
| fallentree | btw be careful with windows and smb... there's that wannacry thing :) | 11:13 |
| ArchaicLord | whats that? | 11:14 |
| ArchaicLord | oh the ransom ware thing | 11:14 |
| fallentree | the ransomware that's been hitting the news for the past few days? | 11:14 |
| fallentree | yah | 11:14 |
| fallentree | make sure you disable smbv1 | 11:14 |
| fallentree | (on the windows side) | 11:14 |
| ArchaicLord | ok will do that... I have just finished universtiy. I aquired a job for a company currenlty using ubuntu as their main os so I am hoping I can fully switch out of windows. BUt I will need to keep my windows instance for a bit | 11:16 |
| ArchaicLord | just in case | 11:16 |
| ArchaicLord | and I think I can' tplay a few games on linux | 11:16 |
| fallentree | ArchaicLord: WINE never ceases to amaze me how smoothly it can run some things :) | 11:20 |
| ArchaicLord | yeah I have dabled with it.. I have never been able to get it to run League of Legends | 11:20 |
| ArchaicLord | which is ultimtly the one game I adore the most | 11:20 |
| ArchaicLord | but agian there is a lot I dont understand about LInux and I can't retain the info either | 11:21 |
| ArchaicLord | I was hoping once I got my ubuntu server up and running hosting my files to look into if i can package up my widnows as is and transfer it into a kvm inside ubuntu | 11:22 |
| ArchaicLord | so for hosting files/ streaming music/dvds what would be ur suggestion? | 11:22 |
| fallentree | suggestion for what? | 11:25 |
| ArchaicLord | server application to host music, films and my own files | 11:32 |
| ArchaicLord | so then i dont need to use spotify, google docs and things | 11:32 |
| fallentree | I wouldn't know what to suggest. | 11:32 |
| fallentree | I don't deal with that kind of services | 11:32 |
| ArchaicLord | no worries thought I would ask incase there is anything new | 11:33 |
| ArchaicLord | well I am pleased I didn't have to trash the raid adn start again :D | 11:33 |
| ArchaicLord | thanks again fallentree | 11:33 |
| aoam | hello | 12:08 |
| aoam | id like to ask, is it possible to set lxd containers to be separated (do not see each other) and see to internet, but not in the way when i setup /etc/network/interfaces, but setup /31 from lxdbr or somehow like that, thanks | 12:09 |
| cpaelzer | aoam: you can create two bridges instead of only the default lxdbr0 and link them up to one or the other - would that suit your needs? | 12:10 |
| aoam | im going to have more containers, maybe .. 80, isnt that problem? | 12:11 |
| cpaelzer | aoam: no problem | 12:11 |
| aoam | so it would setup per conteiner one bridge | 12:11 |
| cpaelzer | aoam: I don't know the limit on bridges thou, but it should work | 12:12 |
| cpaelzer | aoam: essentially you can have a script that sets up a custom bridge the way you want it, and then creates a lxd profile to link it up there to then start that container with that profile | 12:13 |
| cpaelzer | aoam: but since I now read that you want to scale up but nothing see each other wI wondere if there is a better way | 12:14 |
| cpaelzer | aoam: mayb not type bridge at all for the uplink - let me check | 12:14 |
| aoam | i’v found that theres p2p nictype but it doesnt working when i set it up | 12:15 |
| cpaelzer | well you get a virtual dev in the host that you then need to link up right? | 12:16 |
| aoam | p2p: Creates a virtual device pair, putting one side in the container and leaving the other side on the host. | 12:16 |
| aoam | https://github.com/lxc/lxd/blob/master/doc/containers.md | 12:16 |
| cpaelzer | sure "leaving the other side" | 12:16 |
| cpaelzer | I'd more think that macvlan might help - IIRC multiple macvlans don't see each other (only if the switch sends them back) | 12:17 |
| cpaelzer | but that was on s390x OSA cards, other cards might shortcut and reflect the traffic (which usually is good but not for your case) | 12:17 |
| aoam | i’v even tried macvlans but it has the same effect, conteiners cant see each other and also cant see to the internet | 12:18 |
| aoam | so theres the only one possibility, to have per container one bridge and then connect them, thanks :/ | 12:22 |
| cpaelzer | aoam: more experienced container networkers might see a better one | 12:28 |
| cpaelzer | stgraber: ^^ better solutions? | 12:28 |
| cpaelzer | aoam: isn't that a better solution https://serverfault.com/questions/388544/is-it-possible-to-enable-port-isolation-on-linux-bridges ? | 12:30 |
| aoam | wow? it seems that thats working ( ebtables --append FORWARD --logical-in vmbr1 --jump DROP ) | 12:33 |
| aoam | Yup, it works. Thanks | 12:34 |
| aoam | also one more question, i cant find how to set ebtables pernamently after reboot | 12:45 |
| ahasenack | is there an ebtables-save command, like there is for iptables? | 12:46 |
| aoam | yes it is | 12:46 |
| ahasenack | actually | 12:47 |
| ahasenack | aoam: /etc/init.d/ebtables | 12:47 |
| ahasenack | aoam: I *think* that if you call that with "save", it will automatically restore on the next boot | 12:48 |
| ahasenack | that initscript has some interesting options | 12:48 |
| ahasenack | case "$1" in | 12:48 |
| ahasenack | start) | 12:48 |
| ahasenack | [ "$EBTABLES_LOAD_ON_START" = "yes" ] && load | 12:48 |
| ahasenack | you should edit /etc/default/ebtables to your liking | 12:48 |
| aoam | i did /etc/init.d/ebtables save, and then enabled on start, thanks, it works :) | 12:53 |
| aoam | its perfect guys, i’v lost with that a lot of time :) have a nice day | 12:54 |
| ahasenack | \o/ | 12:54 |
| cpaelzer | aoam: yw | 13:10 |
| zetheroo | I have been trying to install Ubuntu 16.04 Server on a system with two identical HDD's in it. I want to setup softraid 1 but grub fails to install every time no matter how I have tried to do the partitioning - which includes following the official documentation https://help.ubuntu.com/lts/serverguide/advanced-installation.html | 13:32 |
| cpaelzer | jamespage: starting to build the OVS 2.7 against the new DPDK that I'm prepping and testing - anything on the OVS changed to be aware of? | 13:32 |
| jamespage | cpaelzer: don't think so - upstream took off the experimental status | 13:32 |
| jamespage | but that's it | 13:32 |
| cpaelzer | thanks jamespage | 13:33 |
| cpaelzer | zetheroo: you install from a server CD like - http://releases.ubuntu.com/16.04.2/ubuntu-16.04.2-server-amd64.iso? | 13:33 |
| mason | zetheroo: Any chance your disks came in formatted with GPT? If so and if you didn't change to legacy MBR, you'd have lacked a uefi_boot partition, which would certainly make GRUB fail. Of course, if you're actually running UEFI, that's a very different situation all by itself. | 13:35 |
| zetheroo | cpaelzer: I installed by following the Ubuntu 16.04 'Ubuntu Server Guide' I linked to. That should work ... or!? | 13:36 |
| ahasenack | zetheroo: it should, I tried that the other day on a VM with two disks and it worked just fine | 13:36 |
| cpaelzer | zetheroo: sure, I just wanted to know from which iso (or whatever) to retry on a VM - also look for mason comment | 13:36 |
| cpaelzer | ahasenack: thanks that lets me skip my test | 13:37 |
| ahasenack | I created / and swap on raid, so in the end I had something like /dev/md0 and /dev/md1 (swap and /) | 13:37 |
| zetheroo | cpaelzer: I am using the iso you linked to, yes | 13:37 |
| ahasenack | zetheroo: someone else had a similar problem here a few weeks ago, in that case the issue was UEFI boot | 13:37 |
| mason | ahasenack: A nifty trick if you're in the pre-ZFS world is to make one big RAID and cut volumes out of LVM sitting atop it. | 13:37 |
| ahasenack | I think he disabled it and enabled legacy boot, then it worked | 13:37 |
| ahasenack | mason: yeah, love lvm | 13:38 |
| zetheroo | ahasenack: ok, will try that | 13:38 |
| ahasenack | I have a mix here | 13:38 |
| zetheroo | what's the benefit of UEFI actually? | 13:40 |
| ahasenack | pass :) | 13:41 |
| zetheroo | ok, well the USB stick (with Ubuntu Server install) is not booting with UEFI and neither are the HDD's | 13:44 |
| zetheroo | I don't know if there is somewhere else that UEFI needs to be disabled from in the BIOS | 13:44 |
| ahasenack | there should be a legacy mode | 13:45 |
| ahasenack | but, I don't have hw with uefi, so I can't tell from experience | 13:45 |
| zetheroo | for each device individually or ....? | 13:45 |
| zetheroo | hmm | 13:45 |
| ahasenack | in general I think | 13:45 |
| dpb1_ | yes, in general | 13:45 |
| mason | zetheroo: Cleaner multibooting, possibility for SecureBoot. | 13:47 |
| dpb1_ | zetheroo: this answer is good: https://askubuntu.com/a/647604/7056 | 13:48 |
| mason | zetheroo: You should have a legacy mode available in any event. | 13:48 |
| zetheroo | I am looking through the BIOS for legacy boot or something similar | 13:48 |
| mason | zetheroo: Also, Ubuntu is fine running inside UEFI if you do want to run in the hardware's preferred mode. You simply have to partition accordingly. | 13:49 |
| mason | zetheroo: Might also be called CSM | 13:49 |
| zetheroo | mason: well that's the thing .. if I do the partitioning how I normally did it before (on the RAID device - Guided use all space) there is automatically an efi boot partition made .. so I thought that would work ... | 13:50 |
| mason | zetheroo: To give you an idea of the possibilities, I'm running on UEFI right now with MD-RAID1 EFI System Partition and ZFS mirrored across a pair of LUKS block devices. | 13:51 |
| mason | zetheroo: Yeah, if you let it partition, it should do the right thing. | 13:51 |
| zetheroo | right, but grub fails to install at the end | 13:51 |
| mason | If you manually partition, you'll want to be intimately aware of the requirements. | 13:51 |
| mason | zetheroo: It fails when you allow it to partition on its own? | 13:51 |
| zetheroo | well this is more automated than the documentation and it still doesn't work | 13:52 |
| zetheroo | yes | 13:52 |
| mason | zetheroo: My recommendation is to use dd if=/dev/zero across both disks. Let it start fresh and add a partitioning scheme, etc. | 13:52 |
| zetheroo | well I can't find any legacy setting on this bios | 13:53 |
| mason | Might be hard to find, but it probably has one. That said, UEFI works fine. | 13:54 |
| zetheroo | mason, is this how you setup your softraid https://help.ubuntu.com/lts/serverguide/advanced-installation.html ? | 14:02 |
| mason | zetheroo: No, I use LVM. | 14:03 |
| zetheroo | ic | 14:03 |
| mason | Well. I use ZFS. But I used LVM before that. | 14:03 |
| zetheroo | One thing I notice is that I cannot change the bootable flag to 'on' | 14:12 |
| zetheroo | I press enter and it remains 'off' | 14:12 |
| mason | That's fine. That's a legacy setting. | 14:15 |
| zetheroo | so leaving 'off' then | 14:15 |
| mason | Should be fine. | 14:15 |
| zetheroo | So this is what I have now http://tinypic.com/r/9ub2w6/9 | 14:21 |
| zetheroo | as per the documentation | 14:21 |
| zetheroo | proceeding with the installation ... | 14:24 |
| zetheroo | grub failed to install :( | 14:47 |
| zetheroo | http://tinypic.com/r/339ocas/9 | 14:48 |
| mason | zetheroo: You need to nail down if you're in UEFI mode or not. | 14:50 |
| zetheroo | any ideas? | 14:50 |
| zetheroo | hmm | 14:50 |
| mason | Your layout is fine for legacy booting | 14:51 |
| mason | But if you're on GPT, it'll fail, and if you're on UEFI, it'll fail. | 14:51 |
| mason | So: 1. dd if=/dev/zero of=yourdisk across both your disks prior to install, as then the installer will Do The Right Thing. 2. Make sure you're explicitly in legacy mode booting, because you'll have to do something quite different if you're booting UEFI. | 14:52 |
| ArchaicLord | can any help me please. I am in UK. My Broadband is supplied by BT. In order to use my own router I have the BT Router set on network 1.254 I had to have my own router set on 0.1 so i beileve its a different subnet. on the 0.1 network I have no a ubuntu web server which I want to host application like next cloud. i am hoping i have set up ddclient to connect to my dynu account to update the ipaddre | 14:53 |
| mason | If you've got GPT partitioning (dunno!) then you'd need to add a bios_grub partition to each disk as well as what you've got, for the combination of legacy booting on GPT. If you've got legacy booting on MBR you don't need this, and the wipe will make that happen. | 14:53 |
| ArchaicLord | eg xxx.xxx.1.254 and xxx.xxx.0.1 | 14:53 |
| mason | If you've got UEFI, then you need GPT and you need an EFI System Partition, but not a bios_grub partition. | 14:54 |
| mason | ArchaicLord: You don't want a separate subnet on a public space. You need NAT. | 14:56 |
| ArchaicLord | mason: ok how and where do I learn to do this properly | 14:58 |
| mason | ArchaicLord: https://help.ubuntu.com/community/Internet/ConnectionSharing maybe | 15:00 |
| zetheroo | mason: it seems I have to change the SATA Mode to IDE (it's currently AHCI) | 15:00 |
| mason | zetheroo: You shouldn't have to do that. I would in fact strongly advise against it. | 15:00 |
| mason | zetheroo: While IDE counts as "legacy" for what it is, it's not the droid you're looking for. | 15:01 |
| zetheroo | mason: SATA Mode Selection This item selects the mode for the installed SATA drives. The options are IDE, AHCI and RAID. SATA RAID Option ROM/UEFI Driver (Available if the item above - SATA Mode Select is set to AHCI or RAID) Select Enabled to use the SATA RAID Option ROM/UEFI driver for system boot. The options are Enabled and Disabled. | 15:01 |
| mason | zetheroo: Maybe take some screenshots of your BIOS. | 15:01 |
| mason | zetheroo: I don't think you want your BIOS doing RAID. | 15:01 |
| ahasenack | +1, don't do that | 15:01 |
| ArchaicLord | mason: my server points to xxxx.xxx.1.254 as its gateway | 15:01 |
| zetheroo | screenshot coming | 15:02 |
| mason | ArchaicLord: Okay. And you likely have a single IP assigned, and that IP is your window onto the world. | 15:02 |
| zetheroo | my motherboard is X10slm-f btw | 15:02 |
| mason | ArchaicLord: Anything behind your firewall will live on a private address space. | 15:02 |
| mason | zetheroo: Doesn't ring a bell. Screenshots FTW. | 15:02 |
| zetheroo | it's a Supermicro board | 15:03 |
| zetheroo | screenshot coming | 15:03 |
| zetheroo | http://tinypic.com/r/mhgi1k/9 | 15:04 |
| zetheroo | disable the 'SATA RAID Option ROM/UEFI Driver' ? | 15:05 |
| compdoc | what a horrible website | 15:06 |
| compdoc | is there a problem with youir drives? | 15:06 |
| zetheroo | compdoc: not that I know of ... why? | 15:08 |
| compdoc | ahci is a good choice, but raid also enables ahci, so Ive heard | 15:08 |
| zetheroo | mason: wdyt? | 15:09 |
| mason | zetheroo: looking | 15:10 |
| zetheroo | k | 15:10 |
| mason | zetheroo: Cab you catch each menu like that? It's going to be something further over to the right. Also, might help to reset to factory defaults. You don't want that BIOS RAID turned on. | 15:11 |
| mason | And that BIOS knows about UEFI, so you'll need to clear that up. | 15:11 |
| zetheroo | mason: I already tried restoring to 'Optimized Defaults' | 15:12 |
| mason | zetheroo: Snag a snapshot of the screen that talks about boot options. | 15:12 |
| zetheroo | ok | 15:13 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_17-14-33-IHys49Gx.png | 15:15 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_17-14-44-PMkToHau.png | 15:15 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_17-14-55-hsoqATaI.png | 15:16 |
| mason | zetheroo: Maybe the security screen too? | 15:17 |
| mason | I'd tend to expect what we want on the Boot screen (with boot options) but maybe it's on Security instead. | 15:18 |
| zetheroo | nothing uefi-related in there ... but just a sec | 15:18 |
| mason | On the plus side, the Boot screen shows the nice variety of UEFI boot manager. | 15:19 |
| Pici | /25/ | 15:19 |
| zetheroo | Under Advanced there is Boot Feature | 15:19 |
| mason | A picture is worth a thousand words. | 15:20 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_17-20-52-8FsoBv3K.png | 15:21 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_17-21-03-c42nZmdG.png | 15:21 |
| zetheroo | this is the motherboard manual https://www.supermicro.com/manuals/motherboard/C222/MNL-1428.pdf | 15:22 |
| mason | zetheroo: Yeah, not seeing it. Welcome to the world of UEFI, which you'll be using with that server! :) | 15:22 |
| zetheroo | but I can't find anything which clearly says how to disable UEFI | 15:22 |
| zetheroo | shit | 15:22 |
| mason | It's not a bad thing. | 15:22 |
| mason | UEFI works fine, and you've got one of the better boot management interfaces. | 15:22 |
| zetheroo | but no documentation :P | 15:23 |
| mason | Wait. | 15:23 |
| mason | So, the Re-try Boot menu should have a legacy mode in it,. | 15:23 |
| mason | try that | 15:23 |
| zetheroo | ha | 15:23 |
| zetheroo | Legacy or EFI boot | 15:24 |
| mason | FWIW, I searched for "legacy" in the PDF you linked. | 15:24 |
| zetheroo | doh | 15:24 |
| mason | If that had turned up dry, the next search would be for "csb". | 15:24 |
| mason | But do read more about UEFI someday. It's not at all bad. | 15:24 |
| mason | Just different. | 15:24 |
| zetheroo | will do ... some day ... | 15:25 |
| zetheroo | so this should do the trick? | 15:25 |
| mason | Leave a tip in the jar. | 15:25 |
| mason | Maybe you'll have other issues, but this seems like a good start. You might still need to wipe the disks to get rid of GPT formatting, if it's there. The installer doesn't know how to deal with that gracefully. | 15:25 |
| dpb1_ | +1 on UEFI, it's worth learning about. it's the way all servers are being built now. | 15:25 |
| dpb1_ | and desktops | 15:26 |
| helpImStuck | is it ok to task questions regarding lxd/lxc here, running ubuntu server as host. | 15:26 |
| helpImStuck | ask :) | 15:26 |
| mason | helpImStuck: Sure. Sadly, I have no experience with lxd, but other folks probably do. | 15:26 |
| zetheroo | well I don't mind using UEFI or whatever, so long as it doesn't mean installing an OS like Ubuntu suddenly takes all kinds of hoops to jump through that were not needed before :P cost vs benefit | 15:27 |
| dpb1_ | helpImStuck: you can, #lxcontainers might be a better community. | 15:27 |
| helpImStuck | dpb1_, thanks :) | 15:28 |
| nacc | how's bug squashing day going (/me just waking up) | 15:31 |
| zetheroo | mason: when formatting the disk should I use GPT or MBR? | 15:37 |
| mason | zetheroo: IIRC the installer doesn't give you an option, but if you're booting legacy, use MBR. | 15:38 |
| mason | This is why I recommended wiping with dd to make sure. But give it a try. | 15:39 |
| zetheroo | I booted into Ubuntu Live and used Disks to format the two disks .. the default was GPT - it seemed to say that GPT was for disks larger than 2TB ... ? | 15:39 |
| mason | zetheroo: Alright, if you're using big disks, you *will* want GPT. The difference between the printed docs and what you want is that you'll format as GPT and have one bios_boot partition per disk. | 15:49 |
| mason | bios_grub partition type is ef02 FWIW | 15:49 |
| zetheroo | is that what that bios_grub thing is? | 15:49 |
| mason | yes | 15:49 |
| mason | You can give it 1MB and that'll be fine. | 15:50 |
| zetheroo | I just don't get why each partition has to be it's own raid | 15:50 |
| mason | per disk, as after install you'll want to make sure both disks are populated - see dpkg-reconfigure grub-pv | 15:50 |
| mason | sorry, dpkg-reconfigure grub-pc | 15:50 |
| mason | Each partition is a raid COMPONENT. | 15:51 |
| zetheroo | why can't you just make one md device and then make all the partitions on the one raid device? | 15:51 |
| mason | And normally you'd want to use LVM. | 15:51 |
| mason | You can and should, but LVM is what lets you do this. | 15:51 |
| zetheroo | and without LVM? | 15:51 |
| mason | In the pre-ZFS world, one must use distinct tools for each layer. | 15:51 |
| zetheroo | well I am trying now like this http://picpaste.com/Screenshot_from_2017-05-17_17-58-58-KS7WpFba.png | 15:59 |
| nacc | Hey everyone! It's Ubuntu Server Bug Squashing Day #4! Planning at: http://pad.ubuntu.com/JxBHprOBVM | 16:00 |
| mason | zetheroo: You might have an issue there. bios_grub needs to be on old metadata if you're doing it as RAID | 16:02 |
| mason | zetheroo: You want to have one bios_grub per disk, not in the raid. | 16:02 |
| mason | You can have it in a RAID, but that's a bit funkier than you might want at present. Ubuntu will handle populating it on two disks automatically, so you might as well let it. | 16:02 |
| zetheroo | gah, I never had this kind of trouble with Raid1 before :P | 16:02 |
| zetheroo | I just let the Guided option make what it wanted to on the Raid device | 16:03 |
| robinwassen | Hi :) | 16:03 |
| mason | The guided option did that? | 16:03 |
| zetheroo | yes | 16:03 |
| mason | Well. Try it then. Interesting. | 16:03 |
| mason | If it blows up, go back to having one bios_grub per disk, outside of the RAID. | 16:04 |
| mason | If it works, please mention that in here. | 16:04 |
| zetheroo | I manually made a single partition on each disk to be used as raid devices, then made the md device using those two partitions, then let the Guided option make those partitions on the md device | 16:04 |
| mason | Ah. | 16:05 |
| mason | I'm betting it blows up then. | 16:05 |
| mason | But we'll see. | 16:05 |
| zetheroo | :D | 16:05 |
| robinwassen | I would like to contribute to solving some bug in Ubuntu, but I don't know where to start. Anyone got a good guide on how the process looks? | 16:07 |
| nacc | robinwassen: have you found a bug? | 16:07 |
| robinwassen | I am a developer, but not very familiar with how the process looks when contributing to Ubuntu | 16:08 |
| dpb1_ | hey robinwassen :) | 16:08 |
| nacc | https://bugs.launchpad.net/ubuntu/ | 16:08 |
| nacc | there are only 131659 to trawl through :) | 16:08 |
| robinwassen | @nacc: https://bugs.launchpad.net/hundredpapercuts/+bug/1646025 I was thinking of this | 16:08 |
| ubottu | Launchpad bug 1646025 in One Hundred Papercuts "libreoffice 5.1.6 for xenial" [Critical,Triaged] | 16:08 |
| nacc | robinwassen: cool | 16:08 |
| robinwassen | Seems easy enough | 16:08 |
| nacc | heh | 16:08 |
| nacc | libreoffice is ... never easy | 16:08 |
| nacc | but this one might be, if it's just a dotrelease | 16:08 |
| robinwassen | Ah, thought it was a stable release of libre and the reference of what version to install as default just had to be updated | 16:09 |
| nacc | robinwassen: no, i think they mean the version packaged in 16.04 needs to be updated | 16:09 |
| nacc | but i see 5.1.6~rc2 is in xenial-updates and xenial-security | 16:10 |
| nacc | mdeslaur: --^ do you know if the security team is going to do a 5.1.6 full release? | 16:10 |
| robinwassen | @nacc: Can you link to where you saw that? :) | 16:10 |
| nacc | robinwassen: the versions? there are two ways: http://pad.lv/u/libreoffice near the bottom and 2) the `rmadison` tool | 16:11 |
| mdeslaur | nacc: rc2 is the final release | 16:12 |
| robinwassen | Thanks | 16:12 |
| nacc | mdeslaur: oh ok, so that bug should be closed then? | 16:12 |
| mdeslaur | nacc: yes | 16:12 |
| nacc | mdeslaur: thanks | 16:12 |
| nacc | robinwassen: you did it! :) | 16:12 |
| robinwassen | @nacc: It solved itself? :) | 16:13 |
| robinwassen | By being included in a patch? :D | 16:13 |
| nacc | robinwassen: well, it's confusing why one task is fix released and one is not | 16:13 |
| nacc | i don't really understand the mgmt of the one hundred papercuts project | 16:13 |
| nacc | dpb1_: do you? | 16:13 |
| dpb1_ | nacc: nope | 16:14 |
| dpb1_ | :/ | 16:14 |
| nacc | dpb1_: sorry, killed my screen :) -- yeah, ok | 16:15 |
| robinwassen | Any chance that you can point me in the right direction of solving some other bug? :) | 16:15 |
| nacc | robinwassen: probably server bitesize is good | 16:15 |
| robinwassen | Like finding a suitable bug and then point to where I find the source repository that contains it | 16:15 |
| ahasenack | https://goo.gl/H66PTQ bite-sized bugs | 16:16 |
| nacc | robinwassen: we can talk about the latter generically as well | 16:16 |
| nacc | ahasenack: thanks! | 16:16 |
| nacc | robinwassen: so given a src package we have a couple of optoins (note that lp bugs are always against src packages) | 16:16 |
| nacc | robinwassen: easiest for now is `pull-lp-source <srcpkgname>` | 16:16 |
| ahasenack | logwatch has some simple ones, but it's a config issue, not something that requires development | 16:17 |
| robinwassen | @ahasenack: The easier the better I guess at this point, I just need to get the workflow to start with, that is the challenging part. | 16:18 |
| zetheroo | mason: check this out ... the layout on the 10th minute https://youtu.be/kfuByWYDlhI | 16:18 |
| ahasenack | right | 16:18 |
| zetheroo | mason: there raid devices ... one of which is for boot | 16:19 |
| zetheroo | three* | 16:19 |
| zetheroo | http://picpaste.com/Screenshot_from_2017-05-17_18-18-03-2tOQeeaV.png | 16:19 |
| mason | zetheroo: That's different, as the BIOS can see something that looks like its regular partition. | 16:20 |
| mason | Not all inside one RAID | 16:20 |
| zetheroo | yeah | 16:20 |
| zetheroo | it's different than the documentation as well | 16:20 |
| mason | zetheroo: Note that he's on MBR partitioning, with smaller disks that don't need GPT. | 16:21 |
| mason | This is a critical difference. | 16:21 |
| zetheroo | how do you know it's MBR though? | 16:21 |
| mason | zetheroo: Size and lack of bios_grub | 16:21 |
| zetheroo | he remakes empty partitions on the disks and is not asked what type he wants ... which is just like here | 16:22 |
| mason | You don't have to believe me, of course. :P | 16:23 |
| zetheroo | well bios_grub is made by the Guided option .. he did everything manually | 16:23 |
| mason | Well, give it a try. Can't hurt to see what happens. | 16:24 |
| nacc | robinwassen: would you like my 7 step (or so) guide to fixing a package? | 16:25 |
| nacc | let me right it a bit more clearly | 16:25 |
| dpb1_ | robinwassen: I would | 16:26 |
| zetheroo | mason: ok, my previous attempt failed :) | 16:26 |
| dpb1_ | er | 16:26 |
| dpb1_ | nacc: I would | 16:26 |
| dpb1_ | :) | 16:26 |
| robinwassen | @nacc: I would love it ! :D | 16:26 |
| zetheroo | mason: so last try here ... going to make a bios_grub on each disk | 16:32 |
| zetheroo | 100MB should be enough ... or? | 16:33 |
| mason | zetheroo: 1MB will be enough. | 16:36 |
| mason | zetheroo: 100 would be a bit of a waste | 16:36 |
| zetheroo | :) | 16:36 |
| mason | Also, you don't need a separate /boot if you're not using encryption. | 16:37 |
| nacc | dpb1_: robinwassen: http://paste.ubuntu.com/24593656/ | 16:39 |
| nacc | i think that's right | 16:39 |
| nacc | robinwassen: we are working on a git-based workflow that wraps some of this up, but even there, 2)-5) will be the same process | 16:39 |
| nacc | robinwassen: just with git commits rather than having to remember what you did :) | 16:39 |
| zetheroo | mason: ok, how does this look? http://picpaste.com/Screenshot_from_2017-05-17_18-40-45-gaBQ9rhK.png | 16:41 |
| patdk-lp | 1MB is normally enough, depends | 16:41 |
| patdk-lp | but you have to be doing something really really odd with grub | 16:41 |
| mason | zetheroo: Looks good from here. | 16:42 |
| zetheroo | k | 16:42 |
| zetheroo | going to give it a try | 16:42 |
| mason | BTW, if you ever actually use all that swap you'll be deeply unhappy with life. :P | 16:42 |
| zetheroo | yeah | 16:43 |
| zetheroo | I don't care at this point :P | 16:43 |
| zetheroo | already so unhappy :D | 16:43 |
| robinwassen | @nacc: Correct me if I am wrong, a normal workflow is to include patches that are applies on packages rather than fixing the problem directly in the package itself? | 16:43 |
| robinwassen | (my interpretation of quilt patches) | 16:44 |
| nacc | robinwassen: that's what `dpkg-source --commit` does | 16:44 |
| nacc | robinwassen: yeah | 16:44 |
| nacc | robinwassen: in some future world (we are working to create) the difference between the two optoins will be managed by tooling | 16:45 |
| nacc | robinwassen: and you can provide us just your changes however you want and we'll figure it out | 16:45 |
| nacc | robinwassen: not sure if that addressed your question or not, i guess | 16:46 |
| robinwassen | I would guess these patches makes it quite a challenge to update versions of the package released by the author :) | 16:46 |
| robinwassen | I think I understand now :) | 16:46 |
| nacc | robinwassen: you mean like the upstream version? | 16:46 |
| robinwassen | yep | 16:46 |
| nacc | robinwassen: right, so one of the steps on every upstream bump is to refresh patches | 16:46 |
| nacc | robinwassen: which often includes dropping them (if fixed upstream) | 16:46 |
| nacc | robinwassen: or, in our case, if debian has picked them up, we don't need to keep them separately ourselves | 16:46 |
| clandest | Hello. I am trying to access my local web server that i have running on my Ubuntu machine from other computers on my network with my Ubuntu's network ip address. I have the server ruinning on 127.0.0.1:8080, i added 192.168.1.22 to my /etc/hosts file but i still cant access the network address even on the ubuntu machine. Would anyone know how i can go about configuring this so that I can access my | 17:01 |
| clandest | local webserver on my network computers? THank you | 17:01 |
| dino82 | nginx or apache | 17:02 |
| nacc | clandest: --^ | 17:02 |
| nacc | clandest: um, if you have the server listening explicitly on 127.0.0.1:8080, then it won't listen on another address | 17:02 |
| clandest | dino82: nacc: tyvm.. I didnt think i would have to use nginx locally, thought there would be an easier way without other programs. but ill try nginx | 17:02 |
| dino82 | Yeah that's your issue, only the machine running the instance will be able to hit it | 17:02 |
| nacc | clandest: you want :8080 to listen on all addresses at that port (iirc) | 17:03 |
| nacc | *all interfaces' addresses | 17:03 |
| clandest | and nginx is prolly the easiest way to achieve that? | 17:03 |
| nacc | clandest: also, dino82 wasn't saying to use nginx, they were asking what you were using? | 17:03 |
| nacc | clandest: no, it's a basic web server decision | 17:03 |
| dino82 | You can use whatever you are comfortable with | 17:04 |
| nacc | clandest: you have told your webserver (based upon what you described) to *only* listen on 127.0.0.1 | 17:04 |
| nacc | clandest: therefore, it doesn't matter about anything else, it will only listen on 127.0.0.1 | 17:04 |
| dino82 | Sorry, I need to be more verbose with my questions, heh | 17:04 |
| nacc | dino82: :) | 17:04 |
| zetheroo | mason: well that seems to have worked | 17:07 |
| zetheroo | grub installed and system is booting | 17:07 |
| mason | zetheroo: \o/ | 17:07 |
| mason | zetheroo: So, the time you spent before was NOT wasted. You learned stuff, and you didn't take it on faith. | 17:08 |
| zetheroo | indeed | 17:08 |
| mason | zetheroo: Now, your next projects are to mess with UEFI and learn ZFS. | 17:08 |
| dpb1_ | robinwassen: part of the bug fixing process in ubuntu is possibly filing a bug on the debian package (if it exists), and then upstream on the source package, if the bug originates there. | 17:09 |
| nacc | dpb1_: oh true, that's a good point, meta tasks that should be at the top | 17:09 |
| dpb1_ | robinwassen: did you find a bug yet to work on? | 17:09 |
| robinwassen | @dpb1: Thanks, it feels like submitting the patch to upstream is at least mandatory :) | 17:09 |
| zetheroo | mason: when ZFS is an option during Ubuntu install .... | 17:09 |
| zetheroo | harhar | 17:09 |
| dpb1_ | robinwassen: usually. unless the bug is in the package!! lol | 17:09 |
| robinwassen | @dpb1_ No luck in finding a bug yet, I am trying to navigate through the huge list! | 17:10 |
| mason | zetheroo: It can be an option (for servers) if you use the graphical installer. | 17:10 |
| robinwassen | Haha | 17:10 |
| mason | zetheroo: It's what I use. | 17:10 |
| dpb1_ | robinwassen: same here actually | 17:10 |
| zetheroo | mason: graphical installer for Ubuntu Server? | 17:10 |
| dpb1_ | ahasenack: do you have pointers? | 17:10 |
| mason | zetheroo: It's the desktop installer, but it gives you access to things like ZFS. You do an install by hand. | 17:10 |
| ahasenack | about what? | 17:10 |
| zetheroo | ok | 17:11 |
| robinwassen | dpb1_: It is easier said than done, and I don't feel like grabbing a bug like "compiz freezes sometimes" :) | 17:11 |
| ahasenack | bugs? I suggested the logwatch ones in the bite-sized list | 17:11 |
| mason | zetheroo: Want to see simple partitioning for a root/boot disk? Here: https://bpaste.net/show/489876e780f2 | 17:11 |
| zetheroo | mason: well thanks for all your help. | 17:11 |
| robinwassen | @ahasenack: Where do I find that one? | 17:11 |
| nacc | robinwassen: only 12 bugs in https://goo.gl/H66PTQ | 17:11 |
| mason | zetheroo: What you don't see from that is that the EFI partition is on RAID, and that root0 is LUKS with half a ZFS mirror inside. | 17:11 |
| mason | zetheroo: My pleasure. | 17:11 |
| nacc | dpb1_: https://goo.gl/H66PTQ | 17:11 |
| ahasenack | https://bugs.launchpad.net/ubuntu/+source/logwatch/+bug/1583705, https://bugs.launchpad.net/ubuntu/+source/logwatch/+bug/1583706, https://bugs.launchpad.net/ubuntu/+source/logwatch/+bug/1578004 | 17:12 |
| nacc | that's the bitesize list | 17:12 |
| ubottu | Launchpad bug 1583705 in logwatch (Ubuntu) "unmatched entries for postfix" [Undecided,New] | 17:12 |
| ubottu | Launchpad bug 1583706 in logwatch (Ubuntu) "unmatched entries for rsyslogd" [Undecided,New] | 17:12 |
| ubottu | Launchpad bug 1578004 in logwatch (Ubuntu) "unmatched entries for courier" [Medium,New] | 17:12 |
| mason | I should have named is esp0. Oh well. | 17:12 |
| mason | s/is/it/ | 17:12 |
| robinwassen | Thanks @ahasenack and @nacc | 17:13 |
| ahasenack | robinwassen: if you are familiar with some service, you could search for bugs against that service too | 17:13 |
| nacc | robinwassen: yeah, that's the other suggestion i'd have, if you have tools you use, or projects you have familiarity with, start there | 17:13 |
| CarlenWhite | I lost understanding of what this script I have that is doing backups to the point I'm looking up into the sky and going, "It's fuckin' magic." | 17:15 |
| CarlenWhite | Script in question. https://pastebin.com/cmkTKaQj | 17:16 |
| Seveas | CarlenWhite: it's doing an unneeded cp, rsync can do this as well using --link-dest. Other than this is a perfectly reasonable simple backup script | 17:18 |
| CarlenWhite | On the page talking about this method mentioned a problem with rsync's --link-dest | 17:19 |
| CarlenWhite | But I'll keep the cp to do what it does best and what rsync does best. | 17:20 |
| CarlenWhite | But I guess I'm pretty tripped up on how hardlinks are being made and how modified files are being done. | 17:21 |
| CarlenWhite | I kinda understand how it might be working when files are deleted. Pretty much hardlink after hardlink is removed until the physical file is no longer reference and marked for free space. | 17:23 |
| CarlenWhite | But for how file modifications are done is screwing with my head. | 17:24 |
| CarlenWhite | Oh! | 17:25 |
| CarlenWhite | rsync! | 17:25 |
| CarlenWhite | It's the one that will create a new inode when a file is modified. | 17:25 |
| CarlenWhite | It all makes sense now. | 17:25 |
| CarlenWhite | Wait does it even do that? | 17:26 |
| dino82 | <3 rsync | 17:26 |
| CarlenWhite | Oh. Does rsync remove a file and then replace it if it needs to update it? | 17:27 |
| CarlenWhite | Because if it's doing that, then it'll pull a new inode to use for the updated file. | 17:27 |
| CarlenWhite | For a moment I thought rsync would zero-byte the file and refill with updated data. | 17:27 |
| andol | CarlenWhite: The default is to rename/replace, but you can also get the overwrite behave by using the --inplace flag | 17:28 |
| CarlenWhite | Which I don't want at all otherwise each instance of myfile.txt in the rotating backup would be replaced with a updated version. | 17:28 |
| CarlenWhite | Assuming myfile.txt was there for the past 7 days and I decided to change something. | 17:29 |
| CarlenWhite | Since myfile.txt from backup.0 to 6 would be sharing the same inode. | 17:30 |
| CarlenWhite | If rsync did a inplace update, it'd update the file with the same inode and ruin previous backups. | 17:30 |
| hallyn | rbasak: (bc i don't see nish here) i would like to formally suggest that server team meeting notes always be pasted in the email in plain text, rather than only having a link. | 17:32 |
| nacc | hallyn: what's up? | 17:33 |
| nacc | hallyn: sure that can be done | 17:33 |
| nacc | we got lazy :) | 17:33 |
| * CarlenWhite quickly updates the backup script with the information so he doesn't have blow his mind again. | 17:33 | |
| CarlenWhite | Quick question, I presume the OS will watch when all hardlinks to the physical location of a file are removed and mark it for free space when it is no longer referenced? | 17:38 |
| hallyn | nacc: oh sorry :) looked for wrong nic :) | 17:51 |
| nacc | hallyn: np | 17:51 |
| nacc | hallyn: i'll bring it up in our team mtg to make sure we do both | 17:52 |
| hallyn | nacc: anyway, i'm probably the only one left reading email with mutt and not wanting to click the link to read the contents :) kernel team does the same thing, | 17:52 |
| hallyn | nacc: awesome, thanks. | 17:52 |
| nacc | hallyn: are you ok if the text is c&p of the link contents? | 17:52 |
| hallyn | yup | 17:52 |
| nacc | hallyn: ok, np | 17:52 |
| mason | Hey, generic question... EL has "yum provides" to identify what package provides something I don't have installed. Is there an equivalent for apt in Xenial and newer? | 17:53 |
| nacc | mason: is provides for package names? | 17:54 |
| mason | nacc: Not package names... Files inside packages. | 17:54 |
| hallyn | mason: dpkg -S /bin/ls | 17:54 |
| nacc | mason: apt-file | 17:54 |
| mason | hallyn: That's only for installed packages. | 17:55 |
| mason | nacc: Looking. | 17:55 |
| nacc | mason: dpkg for installed stuff, apt-file for archive | 17:55 |
| hallyn | mason: yup | 17:55 |
| mason | Is apt-file essentially the same as apt-cache? | 17:56 |
| nacc | mason: no | 17:56 |
| nacc | mason: unrelated beyond both being apt- :) | 17:56 |
| mason | heh, kk | 17:56 |
| nacc | mason: apt-cache searches your apt cache | 17:56 |
| mason | populating now | 17:56 |
| nacc | mason: apt-file searches a package contents list | 17:56 |
| mason | alright | 17:56 |
| mason | I was looking for nslookup earlier, and I found it, but it occurred to me that I didn't know how to search properly. | 17:56 |
| mason | Cool, apt-file works nicely. | 17:57 |
| mason | t | 17:57 |
| nacc | mason: also, command-not-found will tell you the right thing generally for commands; arbitrary files you need apt-file | 17:57 |
| mason | ty* | 17:57 |
| nacc | mason: np | 17:57 |
| mwhahaha | jamespage: sahara-common missing from pike? | 18:33 |
| mwhahaha | jamespage: looks like sahara in general http://logs.openstack.org/70/465670/1/check/gate-puppet-openstack-integration-4-scenario003-tempest-ubuntu-xenial-nv/87bfe0c/console.html#_2017-05-17_17_09_07_591919 | 18:33 |
| === JanC_ is now known as JanC | ||
| Aison | hello | 18:35 |
| Aison | what may be a reason that smbd almost always run at 100% cpu usage | 18:35 |
| ahasenack | Aison: check with smbstatus if it's tied to a particular user | 18:37 |
| Aison | ahasenack, there is no locked file, no user, etc.. | 18:37 |
| ahasenack | Aison: just a lone smbd process at 100%? | 18:38 |
| Aison | yes | 18:38 |
| ahasenack | check the samba logs then, and if they show nothing useful, increase the verbosity | 18:38 |
| nacc | if that doesn't help, Aison, you might strace the smbd process and see what it's actually doing | 18:41 |
| rbasak | hallyn: I'll pass that on, thanks. Nice to know that >0 people care :) | 18:43 |
| hallyn | :) thanks | 18:49 |
| Aison | ahasenack, nacc with log level 4 I don't see anything in the logs | 18:54 |
| ahasenack | then strace it | 18:54 |
| nacc | yeah, i think strace is the next step | 18:54 |
| ahasenack | strace -f -o output -p <pid> | 18:54 |
| ahasenack | something like that | 18:54 |
| ahasenack | leave it a bit, then ctrl-c and inspect the output file | 18:55 |
| ahasenack | Aison: besides the 100% cpu usage, is it working normally? | 18:55 |
| Aison | ahasenack, yes | 18:55 |
| Aison | works normally | 18:55 |
| nacc | if it is pegging the cpu, it'll be quite noisy, but ideally you can see it maybe busy-waiting for a file/lock or something | 18:55 |
| nacc | or determine what loop it's in | 18:55 |
| ahasenack | Aison: wait a sec, with log level 4, you don't see anything *relevant* in the logs, or no logs at all? | 18:56 |
| Aison | I mean, it is a 32core machine, and all cores are somehow in use when samba is running ^^ | 18:56 |
| Aison | that's not normal | 18:56 |
| Aison | ahasenack, there are logs, but they are not growing | 18:56 |
| ahasenack | Aison: maybe there is a config change you are not aware of that is sending them elsewhere? Are the logs you see current? Or from the last time it was restarted? | 18:57 |
| ahasenack | testparm -s can be used to show a dump of the config, with no comments (i.e., it's short) | 18:57 |
| Aison | yes, they are current. I deleted everything in /var/log/samba and then started again. The logs are created | 18:57 |
| Aison | 10 seconds strace creates a 10mb log file | 18:58 |
| Aison | err, 5 seconds | 18:58 |
| ahasenack | :) | 18:59 |
| ahasenack | it's definitely busy | 18:59 |
| ahasenack | you could also try higher log levels perhaps, I don't remember how far up they go | 18:59 |
| ahasenack | but inspect that strace output, look for repeated patterns | 18:59 |
| Aison | that's it: https://people.alvhaus.ch/~ivost/samba.log :) | 19:00 |
| ahasenack | what's /var/log/samba/log.2a02_168_200f_100__2_1, is 2a02_168_200f_100__2_1 an actual machine name? | 19:02 |
| ahasenack | it's "almost" something like 192.168.200.100 :) | 19:02 |
| ahasenack | ipv6 also came to mind :) | 19:03 |
| ahasenack | yeah, ipv6 | 19:03 |
| ahasenack | 2a02:168:200f:100::2 | 19:03 |
| Aison | 2a02:168:200f is my ipv6 prefix | 19:03 |
| ahasenack | that file is opened several times, does it not contain anything useful? | 19:04 |
| Aison | and 2a02:168:200f:100::2 is the IP of the samba server | 19:04 |
| Aison | no, filesize is 0 | 19:05 |
| webnar | hi | 19:06 |
| webnar | someone there? | 19:07 |
| * mason hides in the corner. | 19:07 | |
| nacc | webnar: sure, just ask your question | 19:08 |
| ahasenack | Aison: can you post your smb.conf (output of testparm -s)? Sanitize at will if needed | 19:08 |
| * dpb1_ points at mason | 19:08 | |
| webnar | Ok i'm a basix ubuntu server user so thats why i drop my question here. I had a NAS running ubuntu server 14.04 (was working great). I have upgraded to server 16.04. But now i have issues with my RAID. | 19:09 |
| webnar | Can that be a driver issue or something? | 19:10 |
| webnar | Its working until x hours and then 4 drives get dropped out of the raid. Rebooting fix everything. | 19:10 |
| webnar | ut it are 4 drives connected to my sata controller | 19:10 |
| webnar | the other drives are on the onboard controller | 19:10 |
| nacc | webnar: any messages in the kernel logs when the disks dropped? | 19:11 |
| ahasenack | Aison: do you have something on that server itself connecting to it maybe? I see a ton of accepted connections from itself, if I'm reading that right | 19:11 |
| webnar | no there just gone | 19:11 |
| webnar | also in fdisk | 19:11 |
| nacc | webnar: did you check `dmesg`? disks don't generally silently disappear from the kernel | 19:11 |
| webnar | yes i checked but i really have no clue where to look for. | 19:12 |
| ahasenack | search for the disk device names | 19:12 |
| ahasenack | something like sda usually, but could be sometihng else | 19:12 |
| ahasenack | (in dmesg, that is) | 19:12 |
| nacc | webnar: or pastebin the output | 19:12 |
| nacc | webnar: but it's only relevant if you're in the failed state | 19:13 |
| nacc | if you've rebooted, `dmesg` won't help | 19:13 |
| nacc | you'd need to look in /var/log/syslog or so | 19:13 |
| webnar | i'll post the output when it drops the hd's but still its strange there all on the same controller. | 19:16 |
| ahasenack | webnar: is it software raid? aka, mdadm? | 19:18 |
| webnar | yes | 19:18 |
| ahasenack | webnar: and these are just storage, you have another disk for /, swap, etc? | 19:19 |
| webnar | these are just storage | 19:19 |
| webnar | yes os is on seperate ssd | 19:19 |
| webnar | harddisks show no failures | 19:20 |
| nacc | webnar: right so i'm thinking the driver barfed (maybe) and dropped the controller | 19:21 |
| nacc | webnar: without logs, though, it's hard to know | 19:22 |
| nacc | webnar: did you look at the old logs (or look for them) | 19:22 |
| webnar | where to find old logs? | 19:22 |
| webnar | i only have one dmesg log not a .0 version | 19:26 |
| nacc | webnar: look in syslog | 19:27 |
| webnar | is see dozen of these messages every second of the day in the syslog: | 19:33 |
| webnar | May 17 06:26:12 MediaServer systemd[6908]: dev-disk-by\x2dpartlabel-primary.device: Dev dev-disk-by\x2dpartlabel-primary.device appeared twice with different sysfs paths /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/ata10/host9/target9:0:0/9:0:0:0/block/sdj/sdj1 and /sys/devices/pci0000:00/0000:00:17.0/ata2/host1/target1:0:0/1:0:0:0/block/sdb/sdb3 | 19:33 |
| webnar | May 17 04:09:53 MediaServer mdadm[29886]: Fail event detected on md device /dev/md0, component device /dev/sdj1 | 19:35 |
| webnar | May 17 04:09:53 MediaServer kernel: [141885.926279] md/raid:md0: Disk failure on sdj1, disabling device. | 19:35 |
| webnar | May 17 04:09:53 MediaServer kernel: [141885.928506] md/raid:md0: read error not correctable (sector 1642918560 on sdj1). | 19:35 |
| webnar | May 17 04:09:53 MediaServer kernel: [141886.456458] disk 5, o:0, dev:sdj1 | 19:35 |
| dpb1_ | webnar: can you replace that disk? | 19:36 |
| dpb1_ | it would be my first course of action | 19:36 |
| webnar | its not 1 there 4 disk failing at the same time. | 19:36 |
| webnar | all 4 that are on a seperate sata controller | 19:36 |
| dpb1_ | how many total on the array | 19:36 |
| webnar | 1 of the 4 is even a SSD drive(this one is not in the raid) | 19:37 |
| webnar | 6 total of array | 19:37 |
| dpb1_ | ah, even disks not on the raid, gotcha | 19:37 |
| webnar | 3 on this controller and the other 3 onboard controller | 19:37 |
| dpb1_ | and that controller is internal? | 19:37 |
| webnar | the failing one is a pci controller | 19:37 |
| webnar | But it just happend after upgrading to server 16.04 | 19:38 |
| dpb1_ | what is the controller | 19:38 |
| dpb1_ | lspci should show it | 19:38 |
| webnar | 01:00.0 SATA controller: Marvell Technology Group Ltd. 88SE9230 PCIe SATA 6Gb/s Controller (rev 11) | 19:38 |
| dpb1_ | webnar: you say it's working "fine" for 4 hours | 19:40 |
| === dpb1_ is now known as dpb1 | ||
| dpb1 | do you still get error messages in syslog when it's working fine? | 19:40 |
| webnar | sometimes it can be 4 other times its a few minutes last time it took almost day. | 19:40 |
| webnar | yes | 19:40 |
| webnar | its working now and i get this: | 19:41 |
| webnar | May 17 21:37:53 MediaServer systemd[1]: dev-disk-by\x2dpartlabel-primary.device: Dev dev-disk-by\x2dpartlabel-primary.device appeared twice with different sysfs paths /sys/devices/pci0000:00/0000:00:17.0/ata5/host4/target4:0:0/4:0:0:0/block/sde/sde2 and /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/ata10/host9/target9:0:0/9:0:0:0/block/sdj/sdj1 | 19:41 |
| webnar | there maybe 50 of those on the same second for drive sdj only | 19:42 |
| sarnold | systemd just spews those | 19:42 |
| sarnold | you've got a lot more than I do though | 19:42 |
| dpb1 | webnar: but what about the mdadm and kernel errors. ignore the systemd ones for a sec | 19:43 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.131926] ata8.00: exception Emask 0x0 SAct 0xc SErr 0x0 action 0x6 frozen | 19:47 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.131948] ata8.00: failed command: WRITE FPDMA QUEUED | 19:47 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.131962] ata8.00: cmd 61/08:10:18:10:80/00:00:02:00:00/40 tag 2 ncq 4096 out | 19:47 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.131962] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) | 19:47 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.131993] ata8.00: status: { DRDY } | 19:47 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.132001] ata8.00: failed command: WRITE FPDMA QUEUED | 19:48 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.132014] ata8.00: cmd 61/08:18:38:16:80/00:00:02:00:00/40 tag 3 ncq 4096 out | 19:48 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.132014] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) | 19:48 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.132045] ata8.00: status: { DRDY } | 19:48 |
| webnar | May 17 04:08:31 MediaServer kernel: [141804.132054] ata8: hard resetting link | 19:48 |
| webnar | May 17 04:08:37 MediaServer kernel: [141809.504148] ata8: link is slow to respond, please be patient (ready=0) | 19:48 |
| webnar | May 17 04:08:41 MediaServer kernel: [141814.160288] ata8: COMRESET failed (errno=-16) | 19:48 |
| webnar | May 17 04:08:41 MediaServer kernel: [141814.160342] ata8: hard resetting link | 19:48 |
| webnar | May 17 04:08:47 MediaServer kernel: [141819.524401] ata8: link is slow to respond, please be patient (ready=0) | 19:48 |
| webnar | May 17 04:08:51 MediaServer kernel: [141824.180491] ata8: COMRESET failed (errno=-16) | 19:48 |
| webnar | May 17 04:08:51 MediaServer kernel: [141824.180546] ata8: hard resetting link | 19:48 |
| webnar | May 17 04:08:57 MediaServer kernel: [141829.548697] ata8: link is slow to respond, please be patient (ready=0) | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.108908] ata10.00: exception Emask 0x0 SAct 0xe0000 SErr 0x0 action 0x6 frozen | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.108986] ata10.00: failed command: READ FPDMA QUEUED | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109042] ata10.00: cmd 60/00:88:00:30:39/04:00:0b:00:00/40 tag 17 ncq 524288 in | 19:48 |
| mason | webnar: Maybe try bpaste.net | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109042] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109169] ata10.00: status: { DRDY } | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109206] ata10.00: failed command: READ FPDMA QUEUED | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109258] ata10.00: cmd 60/00:90:00:f4:ec/08:00:61:00:00/40 tag 18 ncq 1048576 in | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109258] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109386] ata10.00: status: { DRDY } | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109421] ata10.00: failed command: READ FPDMA QUEUED | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109473] ata10.00: cmd 60/a0:98:00:fc:ec/06:00:61:00:00/40 tag 19 ncq 868352 in | 19:48 |
| ahasenack | webnar: might be silly, but you should check for loose sata cables | 19:48 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109473] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) | 19:49 |
| mason | webnar: This kind of paste often results in a temporary kick for flooding. | 19:49 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109600] ata10.00: status: { DRDY } | 19:49 |
| webnar | May 17 04:09:02 MediaServer kernel: [141835.109639] ata10: hard resetting link | 19:49 |
| webnar | so i guess it started to stop working from that point | 19:49 |
| webnar | before that there are no erros | 19:49 |
| webnar | and from that moment its full of those errors | 19:49 |
| webnar | yes tried the sata cables and power. But the only thing i noticed was the fact the 4 failing drives are on the pci sata controller | 19:49 |
| dpb1 | webnar: if it were me, my next course of action would be to replace the controller | 19:50 |
| ahasenack | yeah | 19:50 |
| ahasenack | all drives failing at the same time doesn't happen, unless they are rebuilding the array (that causes extra stress) | 19:50 |
| dpb1 | you've narrowed it down pretty well, I'd say. | 19:50 |
| webnar | yes but wouldn't it be something with ubuntu 16.04 because it only happend after the upgrade | 19:51 |
| webnar | https://bpaste.net/show/a66f06183190 | 19:53 |
| webnar | after reboot everything works again | 19:54 |
| ahasenack | webnar: if you think it's some driver, you could boot the previous ubuntu with a live-cd, mount the array and leave it be for a while, see if it also encounters the same problem | 19:58 |
| ahasenack | or install a newer kernel in 16.04 using the hwe series | 19:59 |
| ahasenack | but so far all points at a hardware problem | 19:59 |
| dpb1 | webnar: you might find this thread interesting... https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=700975 | 19:59 |
| ubottu | Debian bug 700975 in src:linux "linux-image-3.7-trunk-amd64: Marvell 88SE9230: Freaks out and drops all disks if sent SMART command during RAID rebuild" [Normal,Open] | 19:59 |
| dpb1 | webnar: I know it's old | 20:00 |
| genii | Hm, 3.7 | 20:00 |
| dpb1 | end result was a recommendation against that controller, and a list of ones with better kernel support: http://blog.zorinaq.com/from-32-to-2-ports-ideal-satasas-controllers-for-zfs-linux-md-ra/ | 20:03 |
| sarnold | oh man I spent -days- on that blog post :) | 20:04 |
| webnar | But the recommended controllers most of them are very outdated. | 20:07 |
| * dpb1 nods | 20:09 | |
| sarnold | it is from 2012 or something. | 20:12 |
| === Mikee_C_ is now known as Mikee_C | ||
| webnar | Hmm i'll try to install 14.04 first and check if it still works with 14.04 like it always did | 20:20 |
| * ppetraki catching up on sata errors | 20:45 | |
| ppetraki | webnar, so umm, how old are these ssds? | 20:47 |
| webnar | few months | 20:47 |
| webnar | just installed server 14.04 | 20:47 |
| ppetraki | webnar, you haven't written like 50TiB of data in the meantime have you? | 20:47 |
| * ppetraki suspects not | 20:48 | |
| bindi | why 14.04 | 20:48 |
| webnar | 14.04 was working perfectly before the issues | 20:48 |
| Aison | ahasenack, [2017/05/17 22:48:09.606627, 4] ../source3/smbd/sec_ctx.c:321(set_sec_ctx_internal) | 20:49 |
| Aison | this is the log entry that is written endless | 20:49 |
| webnar | if i check loggin now there are no errors | 20:49 |
| ahasenack | Aison: is that part of a panic? | 20:49 |
| Aison | so samba stays at 100% cpu usage | 20:49 |
| ppetraki | webnar, scsi error handler is pretty thorough. if it can't recover a drive something is really wrong | 20:50 |
| Aison | ahasenack, no idea :-( these entries are repeated: https://pastebin.com/GJ50YHu8 | 20:50 |
| webnar | yes but before i got the 50 errors a sec on ubuntu 16.04 just installed 14.04 now and no errors in my syslog anymore. | 20:50 |
| ahasenack | Aison: what was the logging level before, 0? Or has it always been at 4? | 20:51 |
| webnar | raid is up and clean | 20:51 |
| ahasenack | webnar: that could also be because of the reboot, right? | 20:51 |
| ppetraki | marvell controller, ok not my favorite but this is a well travelled solution. | 20:51 |
| ahasenack | if it stays like that for a few hours, under load/usage, then yes, this is good info | 20:51 |
| Aison | ahasenack, before it was 0 (testparm always removed log level) | 20:52 |
| Aison | now it is 4 | 20:52 |
| ppetraki | webnar, it's possible that your combination of drive and controller is making a bug pop up in 16.04 | 20:52 |
| ppetraki | webnar, those are hard errors being reported | 20:52 |
| ppetraki | webnar, that or you got hit with a power spike and everything is compromised | 20:53 |
| ahasenack | Aison: maybe 4 is too much detail already | 20:53 |
| Aison | with log level 10 over 100mb/s of log file is written :P | 20:54 |
| ahasenack | yeah, stay away from that :) | 20:54 |
| webnar | I'll post a update on the status after some time. i'll let it run for now. Powerspike i don't think the NAS is on a UPS. | 20:54 |
| ahasenack | although that shows that connections to the server are being made | 20:54 |
| ahasenack | try 3, then 2, 1 | 20:55 |
| ppetraki | webnar, yeah I'm catching up on the rest of your logs | 20:55 |
| ahasenack | and check smbstatus again | 20:55 |
| ppetraki | webnar, I think you found a bug, stuff like this is so simple it shouldnt even happen | 20:55 |
| ppetraki | webnar, you on amd or arm? | 20:56 |
| Aison | ahasenack, here with log level 3 :) https://pastebin.com/aeFdYEf5 | 20:56 |
| * ppetraki nm amd | 20:56 | |
| ppetraki | yeah, weird | 20:56 |
| webnar | intel | 20:56 |
| ppetraki | so... intel cpu and marvel sata controller? | 20:57 |
| webnar | yes | 20:57 |
| ppetraki | onboard? | 20:57 |
| webnar | pci | 20:57 |
| ahasenack | Aison: still doesn't ring a bell | 20:57 |
| ppetraki | oh ok | 20:57 |
| ppetraki | could you move these the onboard and try 16.04 again? that would tell us for sure its a driver issue | 20:58 |
| webnar | all other drives are onboard. and the half of the drives from the raid | 20:59 |
| ppetraki | oh ok | 21:00 |
| webnar | the 3 onboard drives keep running. The 3 PCI sata controlled drives stop. There is a fourth harddisk on the controller download SSD drive | 21:00 |
| webnar | that SSD also stops working. | 21:00 |
| webnar | there is no raid or something on the SSD | 21:00 |
| Aison | ahasenack, maybe it is related to ldap? | 21:00 |
| ahasenack | well, talk about a wrench in the gears | 21:01 |
| ppetraki | webnar, yeah this is just dumb simple io having a bad day. most basic functionality | 21:01 |
| ahasenack | Aison: did you check the pure smbd logs as well, or just the log.<machine> files? There should be a log.smbd file too iirc | 21:02 |
| ppetraki | webnar, this marvell thing is finicky. you can file a bug ... or dump it with a basic lsi or atto and move on with life | 21:06 |
| ppetraki | webnar, you can try reducing the link speed on the driver down to 3G in the hopes it will reduce the probability of this event. which will also reduce you to about 250MB/s. | 21:08 |
| ppetraki | webnar, other than that I don't have any quick fixes for you. sorry :( | 21:08 |
| webnar | anyway thanks for the help. | 21:10 |
| webnar | sudo make | 21:10 |
| webnar | srry wrong screen | 21:10 |
| ppetraki | :), np | 21:11 |
| === god is now known as Guest79555 | ||
| nacc | smoser: around? | 23:22 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
