| === not_phunyguy is now known as phunyguy | ||
| calcmandan-tab | i have an severely out of date server on 15.04 and looking for directions of updating severely eol ersions. can someone kindly send me a url? | 07:11 |
|---|---|---|
| lotuspsychje | calcmandan-tab: is that a test server or production? | 07:54 |
| calcmandan-tab | foud it | 08:03 |
| calcmandan-tab | found it | 08:03 |
| calcmandan-tab | fingers crossed i don't hose my server | 08:03 |
| lotuspsychje | calcmandan-tab: a lot of security flaws have been released since 15.04 are you sure you take the risk of eolupgrade? | 08:04 |
| calcmandan-tab | lotuspsychje: not sure what the risk is | 08:05 |
| lotuspsychje | calcmandan-tab: were your servers connected to internet or offline use? | 08:06 |
| calcmandan-tab | both | 08:06 |
| calcmandan-tab | but not critical | 08:06 |
| lotuspsychje | !usn | calcmandan-tab security risks | 08:06 |
| ubottu | calcmandan-tab security risks: Please see https://usn.ubuntu.com/ for information about recent Ubuntu security updates. | 08:06 |
| lotuspsychje | calcmandan-tab: if its for production use, id really reccomend fresh install to be sure your server isnt compromized | 08:07 |
| lotuspsychje | maybe its not, but why take the risk right? | 08:07 |
| calcmandan-tab | i did a major scan of the server and i have strong egress rules on the firewall | 08:07 |
| calcmandan-tab | i also reviewed its logs | 08:07 |
| calcmandan-tab | firewall and proxy logs | 08:08 |
| lotuspsychje | calcmandan-tab: ok then if you know what youre doing :p | 08:08 |
| cpaelzer | jamespage: did your issue with OVS 2.12 resolve? | 08:32 |
| Aison | hello, can I disable the network-manager.service, when I do not use the /etc/network/interfaces facility? is it enough to keep just systemd-networkd.service enabled? | 09:11 |
| lordievader | Aison: You mention three different methods of configuring network interfaces here. What do you actually use? | 09:24 |
| calcmandan-tab | lotuspsychje: thanks for your help earlier. | 09:24 |
| lotuspsychje | calcmandan-tab: welcome mate | 09:25 |
| calcmandan-tab | lotuspsychje: i'm taking your advice as it'll be much quicker to go fresh | 09:25 |
| calcmandan-tab | lotuspsychje: and cleaner | 09:25 |
| lotuspsychje | +1 nice choice calcmandan-tab | 09:25 |
| calcmandan-tab | night | 09:26 |
| calcmandan-tab | take care | 09:26 |
| Aison | lordievader, i'm only using the systemd facility | 09:26 |
| lordievader | In that case NetworkManager and ifupdown (`/etc/network/interfaces`) can safely be disabled. | 09:27 |
| Aison | so I placed config files into /etc/systemd/network | 09:27 |
| lordievader | Do make sure the `systemd-networkd` service is enabled and running. | 09:27 |
| lordievader | If you are able to, reboot at least once to check if the configuration is reboot-proof. | 09:28 |
| Aison | ok | 09:28 |
| Aison | I just wonder, why networkmanager replaced my reslov.conf by an empty file (except containing # Generated by NetworkManager) | 09:28 |
| lordievader | Because it likes to control it with information either from DHCP or manually supplied. | 09:29 |
| lordievader | If you are doing this remote, make sure you have some backup method of access (IPMI). | 09:29 |
| Aison | the funny thing is, altough resolv.conf has no nameserver defined, nslookup, dig, etc... is still working. | 09:31 |
| Aison | is automatically the localhost used if there is no nameserver defined? because i'm running bind on this machine | 09:31 |
| lordievader | The DNS was designed this way. | 09:31 |
| lordievader | As a way to push over people to the DNS (instead of the hosts file) the fall-back default would be localhost. | 09:32 |
| Aison | lordievader, ok, thx :-) /etc/network/interfaces is read by the network-manager.service or are these two different facilities? | 09:35 |
| lordievader | NetworkManager only reads that file to figure out which interfaces are managed by ifupdown, furthermore it does its own thing. NetworkManager is not configured through `/etc/network/interfaces`. | 09:36 |
| Aison | brb, reboot server :-D | 09:44 |
| Aison | lordievader, ok, everything works fine without network-manager.service ^^ | 09:47 |
| lordievader | Good to hear | 10:02 |
| Aison | argh, why is mariadb logging everything... just deleted a 500gb log file | 10:59 |
| Aison | oh, somehow general-log was enabled.... | 11:03 |
| === Wryhder is now known as Lucas_Gray | ||
| hallyn | cpaelzer: yeah, so qemu supported virtio-scsi-blk, but the eoan server livecd couldn't find the disk :( | 13:32 |
| hallyn | i'm being lazy - haven't looked at the available kernel modules yet | 13:32 |
| hallyn | jamespage: cpaelzer mentioned that rharper had suggested you switch to virtio-scsi, did you do so, and did you have any troubles? | 13:33 |
| hallyn | oh you know what - never mind, | 13:44 |
| hallyn | i using the $(*&^&(*#&$(*#&$(* "improved" qemu command line wrong | 13:44 |
| hallyn | man i hate the -device crap. and now, apparently, you have to use *3* pairs of arguments for one stinking drive. lovely | 13:45 |
| hallyn | but, it's working :) now to test discard | 13:45 |
| cpaelzer | IMHO the cmdline no more is meant to be a human interface | 13:46 |
| cpaelzer | more for machine generated strings | 13:46 |
| cpaelzer | I happen to use the most outdated (but easy to remember) arguments as well every now and then | 13:47 |
| hallyn | yeah... | 13:47 |
| hallyn | i bet there's a front-end out there that parses easier to remember cmdline to qemu cmdline :) maybe i'll turn my 'kvm' script into one | 13:49 |
| hallyn | "-virtio-scsi-drive x.qcow2" -> if=none,id=hd,file=x.qcow2,format=qcow2 -device virtio-scsi-pci,id=scsi --enable-kvm -device scsi-hd,drive=hd | 13:50 |
| hallyn | meh | 13:50 |
| rharper | hallyn: just wait till you start allocating io-threads to each of your devices =) | 14:23 |
| hallyn | rharper: i've always assumed this was all jsut an attempt to force us to use libvirt :) | 14:30 |
| rharper | hallyn: heh, no, that was when they stopped allowing folks to query features via the cli; wanted folks to use qmp to "discover" features; | 14:35 |
| rharper | https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1716028 | 14:36 |
| ubottu | Launchpad bug 1716028 in libvirt (Ubuntu) "qemu 2.10 locks images with no feature flag" [Medium,Fix released] | 14:36 |
| hallyn | half-lol :) | 14:36 |
| === Tornevall is now known as Guest64769 | ||
| hallyn | rharper: so - all that, and discard is not working for me! hm. | 17:19 |
| hallyn | ok, so i see - it doesn't shrink the file of hte qcow file, but it re-uses the discarded space (so create 1G, rm 1G, create 1g doesn't take up 2 G) | 17:29 |
| hallyn | "could be worse" | 17:29 |
| sarnold | it could also be better.. if those bits on disk really aren't needed it'd be nice to zero em out for rebalancing :( | 17:30 |
| lordcirth | sarnold, as a one-off, you can use qemu-img convert to produce a shrunk file | 17:40 |
| sarnold | lordcirth: yeah.. bummer that it takes that kind of effort though :( if the vm actually went to the effort to issue a discard call, that information is already sitting there.. | 17:41 |
| rharper | hallyn: what's your cli look like ? | 17:52 |
| hallyn | kvm -drive if=none,id=hd,file=x.qcow2,format=qcow2 -device virtio-scsi-pci,id=scsi --enable-kvm -device scsi-hd,drive=hd -m 4096 -drive if=none,id=hd2,file=y.qcow2,format=qcow2 -device scsi-hd,drive=hd2 | 17:53 |
| hallyn | rharper: ^ | 17:53 |
| compdoc | I started using raw | 17:56 |
| rharper | you need: -drive if=none,id-hd,file=x.qcow2,format=qcow2,discard=unmap,detect-zeroes=unmap | 17:57 |
| rharper | and you're running fstrim in guest and checking on host afterward ? | 17:57 |
| rharper | well, at least the discard=unmap; I've not yet tested the detect-zeroes either, but I suspect that may be helpful for guests which zero block space | 17:58 |
| hallyn | i thought about detect-zeroes, but my reading of the manpages suggested that all of that would default to what i wanted | 18:01 |
| hallyn | oh, maybe not - maybe htat was seen as too much perf impact? | 18:01 |
| hallyn | yeah, running fstrim in guest then checking host | 18:01 |
| rharper | hallyn: which version of qemu ? | 18:07 |
| rharper | hallyn: I would try being explicit; I see some references to enabling discard by default, but the man pages don't indicate the default discard mode on -blockdev or -drive, or if the defaults are the same; =( | 18:12 |
| hallyn | rharper: QEMU emulator version 4.0.0 (Debian 1:4.0+dfsg-0ubuntu5) | 18:32 |
| hallyn | i added the discard and zeroing options, same behavior | 18:33 |
| hallyn | maybe i needed to wait longer? :) that would make sense | 18:33 |
| hallyn | i assume detecting zeroes takes some time | 18:34 |
| rharper | hallyn: I don't think so; it should be immediate, fstrim in guest to host should result in a punchole syscall to the file on the host ; have to tried raw vs qcow2? | 18:41 |
| hallyn | no | 18:43 |
| hallyn | i suppose this could be qcow2 bogosity | 18:44 |
| * rharper tries it out locally | 18:46 | |
| hallyn | rharper: yes! raw works | 18:54 |
| hallyn | at least, with simple ext4, lemme try lvm now | 18:54 |
| hallyn | rharper: yeah, even lvm thinpools shrink that way. | 19:02 |
| hallyn | whodathunk raw format works better for shrinking size than qcow2 :) | 19:03 |
| hallyn | (we all know - you would :) | 19:03 |
| rharper | hallyn: it's working for me (xenial qemu 2.x) on both qcow2 and raw; | 19:03 |
| hallyn | thanks rharper | 19:03 |
| hallyn | hm | 19:03 |
| rharper | ext4 on both filesystems | 19:03 |
| rharper | lemme repeat on qemu 4.x | 19:03 |
| hallyn | qcow2 doesn't work for me, but that's probably ok | 19:03 |
| rharper | how is your qcow2 built ? base image or layered with a backing file ? | 19:04 |
| hallyn | just a base qemu-img create -f qcow2 x.qcow2 10G | 19:04 |
| rharper | yeah | 19:04 |
| rharper | same here in my tests | 19:04 |
| rharper | hallyn: so, on qemu 4.0 I see almost *no* trim in the qcow2; so I suspect there's some additional flags that need set on the qcow2 -drive parameter | 19:20 |
| rharper | so that confirms what you see | 19:20 |
| hallyn | in some circles that would be called a regression :) | 21:41 |
| keithzg[m] | Hmm, one VM at work (running on KVM) now fails to boot with 4.15.0-58-generic. Still fine with 4.15.0-38-generic, and even more weirdly, other 32-bit VMs using 4.15.0-58-generic boot just fine on the same host. | 22:55 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!