=== JanC is now known as Guest19389 === JanC_ is now known as JanC [01:47] metaf5_: ok, cool [03:43] Confirming that I'm actually an idiot. Apparently cloud-init in 14.04 didn't care that I was capturing an AMI from a once-booted machine, and would happily run the user-data script, but I think in 16.04 the record of the previous boot persisted and makes it not run the once-only cloud-init user-data-script. [03:44] Personally, I solved *my* problem by just adding "#cloud-boothook" to my script. [03:45] It would be lovely if the documentation hinted at this fact though, as I imagine it's a common usecase to boot machines, provision them, capture AMIs, and then provide user-data when booting those AMIs for later configuration. [03:45] I assume this is all expected behavior though. [07:15] hi ,i need create user for ssh ,i want this user access all linux except one directory [07:51] hi ,i need create user for ssh ,i want this user access all linux except one directory [07:52] !acl [07:52] bah [08:54] hi ,i need create user for ssh ,i want this user access all linux except one directory === iberezovskiy|off is now known as iberezovskiy [10:25] cpaelzer: do you know about "git notes"? [10:25] They might be quite useful for your proposal. [10:31] Hi! Does anyone here have experience with performance evaluation and tweaking of dm-cache? [10:32] ➜ ~ sudo dmsetup message vg_cache-origin_device 0 sequential_threshold 128 [10:32] device-mapper: message ioctl on vg_cache-origin_device failed: Invalid argument [10:32] Command failed [10:32] I'm getting this so I can't make any changes to the default policy [10:37] rbasak: I don't but it sounds just right [10:37] * cpaelzer reading [10:42] cpaelzer: you mean the error message I'm getting sounds just right (ie., I'm requesting it wrong)? Or that my command sounds just right and it's strange I got the error message? [10:42] sat_: no I only referred to what rbasak said before [10:42] cpaelzer: oh, sorry, right [10:43] sat_: but actually yours is interesting, reding that as well [10:43] rbasak: I like git notes, less explicit "merge-notes" than I thought but structured by design and with tool support [10:43] rbasak: thanks for pointing that out [10:45] cpaelzer: you can configure git to carry notes through rebases, and git log (and cgit) displays them. But they are tied to commits, rather than being an extra blank commit. We could attach a note to the commit that is tagged with the upload tag, however. [10:45] That way we can write up "upload notes". [10:45] rbasak: yeah [10:45] Or, we could use annotated upload tags, and put them in the annotation. [10:45] notes are "less attached" in the sense that they can be changed and removed at any time. [10:46] Tag annotations are more permanent. [10:46] So there are a few options. [10:47] rbasak: I'm updated my proposal, if you could add on "can configure git to carry notes through rebases" that would be great [10:47] rbasak: I'd also like the annotated upload tag [10:47] rbasak: actually I'm open to whatever carries that message as long as we have a defined way to store it [10:48] rbasak: bonus for any hardening to not get easily lost on rebases, merges and so on [10:58] sat_: I could only think of vg_cache-origin_device not being the right target [10:59] sat_: is that what you created with dmsetup create ... ? [10:59] sat_: but in general the tuning of dm_cache as nice as it is is also its weakness [10:59] [69089.568011] device-mapper: cache: bad config value for random_threshold: 128 [10:59] this is what dmesg gives me [11:00] sat_: maybe value too small? [11:00] I need to check [11:00] I created with lvcreate and lvconvert, that's the way I found on some web pages [11:01] I just checked the source code - there's no sanity checking [11:01] the only thing I can think of now is that the setup isn't as it should be [11:01] sat_: can you set values closer to the default like 512 or 1024 that way ? [11:02] cpaelzer: same message - [69388.271654] device-mapper: cache: bad config value for sequential_threshold: 1024 [11:02] sat_: hrm :-/ [11:02] sat_: I must admit that the tunability of dm-cache is its power and its pain - I mostly seen people more happy with bcache [11:02] so dm knows it's cache, it seems to be almost there... but it's not working [11:03] bcache performance sux - I spent a week evaluating it in writeback mode [11:03] surely depends on the case, if it is not working for you ok [11:03] just tried Intel CAS, it works beautifully, performance is great in writeback mode, but costs [11:03] so wanted to see dm-cache [11:04] bcache could be great as a read cache [11:05] needs more work to be great as a write(back) cache [11:05] sat_: just checked but none of my HW around atm qualifies to try reproducing your case :-/ [11:05] (we're not so interested in the read-cache side) [11:06] cpaelzer: meaning you don't have dm-cache setups? [11:06] sat_: not atm [11:06] ok, I hope some good soul will join the discussion [11:07] I'm not an IRC users normally - should I try to get attention from some other users here? [11:07] IRC users -> IRC user [11:07] cpaelzer, could you recommend someone? [11:08] sat_: if you happen to know people related you would highlight them - but I don't know who to recommend atm [11:08] ok, great, thanks cpaelzer! [11:08] sat_: another way is to describe your case on askubunut - IMHO the community cares more there [11:08] sat_: while devs are more here [11:08] sat_: but that is only my personal view of things [11:09] sat_: if you could link a particular howto you found to be best to follow to set it up that would be nice if I or anybody else find some time to look at it [11:09] * cpaelzer is loving performance stuff [12:34] nacc: how about s/Publish parent/Pocket parent/? More clear to those not familiar with the Launchpad API maybe? [12:34] * rbasak isn't sure. [12:38] Though pockets don't have a meaning in Debian. Perhaps "Series"? [12:38] "Series parent" [12:39] Also, if following git convention, something like "Importer-Series-Parent: ..."? [12:39] And "Importer-Changelog-Parent:" [12:40] Or in fact, non-capitalised "Importer-series-parent: ..." etc. [12:40] Whatever makes sense in the git world :) [12:41] Finally, if doing that, how about "LP: #XXX" if bugs were fixed, parsed from the changelog? [12:41] (must use the standard parser of course, etc) [12:52] rbasak: is the LP: XXX idea to spread tags accordingly on import? [12:53] cpaelzer: I'm not sure what you mean by "spread tags". I just mean to put the references in the commit messages if we know about them. That's a standard format we're using in MPs for buglinks, too. [12:54] rbasak: ah in the commit message, ok [12:54] rbasak: with "spread" I meant to add a tag like release-lp-#### on a commit fixing a bug [12:54] Oh, I see. [12:54] No, I wasn't suggesting that. [12:55] ok [12:55] I'm good reading the changelog and picking the right import [12:55] cpaelzer: Here's the easiest dm-cache setup guide I found: https://videos.cdn.redhat.com/summit2015/presentations/17856_getting-the-most-out-of-your-nvme-ssd.pdf [12:55] I don't have a good use case for the LP: #XXX suggestion. I just thought it'd be nice to be consistent with what Launchpad reads later. [12:56] As the metadata is there. Though only sort of - it can be inferred from debian/changelog, but really it's in the changes file which I'm not sure we have. [12:56] sat_: thanks - I hope to find some time later one [12:56] -e [13:27] hi guys, i got a postfix server running on my ubuntu server. now whenever a new mail arrives in a mailbox i want a notification in my terminal the next time i enter a command, how can I achieve that? [13:28] the "you have new mail" message is not triggering even if i have new mail [13:28] only when relogging the "you have new mail" appears [13:37] macskay_: you'd need to add a dynamic command into the bash PS1 prompt I think [13:38] hi all [13:38] Note: user in #ubuntu jophish has discovered 16.10 server amd64 installer is misisng the iwlmvm kernel module [13:38] The iwlmvm module doesn't seem to be on the disk for server-16.10-x86_64 [13:39] snap :) [13:39] :) [13:39] Is there something I can do to workaround this for the time being? [13:39] jophish: you could copy the module over from a USB device from the desktop install, as a workaround, if it is needed right now [13:40] into /lib/modules/4...generic? [13:40] jophish: yes, then do "depmod" to refresh the dependency list before using modprobe again [13:41] coreycb: i dont know if you knew about this https://wiki.debian.org/Python/LibraryStyleGuide [13:41] TJ-: will I have to run depmod after booting the live disk? [13:42] zul, i think i've seen it but could always use a refresher, thanks [13:43] jophish: I've been assuming you currently have a root shell on the server installer. in which case you can copy it from a (mounted) USB device - even from the ISO image - and do it all without any rebooting [13:43] coreycb: i got new releases [13:43] zul, new dependencies? [13:43] TJ-: I've only got one usb drive with me at the moment [13:44] so I'm going to copy it on from another machine [13:44] jophish: ahhh, I see. OK, well you know what you need so shout me if you need further suggestions [13:44] Thanks a lot for all the help TJ-, I'll let you know how it goes [13:45] coreycb: newer sarahaclient [13:45] zul, ok thanks [13:59] TJ-: loading iwlmvm leads to a who bunch of unknown symbols, ieee80211... [13:59] I've noticed that kernel/drivers/net/wireless is absent entirely [13:59] I guess it's not often that people install ubuntu server on a wireless only machine [14:01] jophish: missing symbols mean you have the module from a different version of the kernel [14:02] jophish: you need to ensure they're exactly the same. tell me what "uname -r" reports on the server installer and I'll find you a link to the file you need [14:02] TJ-: 4.8.0.22-generic [14:03] I copied them from the 4.8.0.22-generic directory on the ubuntu desktop 16.10 iso [14:03] it* [14:03] jophish: in which case the symbols would match :s [14:04] TJ-: I tried again but this time I copied the whole wireless directory [14:04] only two missing symbols now: ieee80211_channel_to_frequency and ieee80211_hdrlen [14:05] TJ-: jophish: Wouldn't it make more sense to install the package that contains the wifi drivers? [14:05] the files would be from/put in "/lib/modules/$(uname -r)/kernel/drivers/net/wireless/intel/iwlwifi/" including the ./mvm/ sub-dir [14:05] Odd_Bloke: it's the 16.10 server installer itself, so limited environment and no network right now [14:05] TJ-: Sure, but you're getting files on there somehow. [14:06] Odd_Bloke: installer kernel is missing the iwlmvm.ko module [14:06] Odd_Bloke: there isn't dpkg there [14:06] Oh, well, that's a good argument then. :) [14:06] Odd_Bloke: :~) [14:06] jophish: the donor is the same arch (amd64), yes? [14:07] yup [14:08] TJ-: to be clear, I got lots of unknown symbols when I had just copied the iwlmvm.ko driver, but many fewer unknown symbols when copying the whole 'wireless' directry [14:08] could I just need to copy some more? [14:10] jophish: unknown symbols really does mean the kernel versions are different. REMEMBER the *installers* kernel version isn't necessarily the same one as is being installed (at least I recall being caught out like that once!) [14:10] jophish: so ensure it is correct in a root shell of the installer with "uname -r" [14:11] jophish: sometimes the installer has an older point release than the one being installed (due to updated ISOs) [14:14] TJ-: that must have been it [14:14] perhaps it's by design that there are no wireless drivers on the server iso [14:14] jophish: no, they are there, you had iwlwifi loaded, but its' sub-module iwlmvm was missed off [14:14] that's a bug [14:14] TJ-: I've just been given a usb disk large enough for ubuntu-desktop to fit on, so I might try installing that [14:15] jophish: tell me what the installer's "uanme -r" reports I'll find you the iwlmvm.ko for that [14:15] ok, one sec [14:15] "uname -r" grr typo [14:15] TJ-: 4.8.0.22-generic [14:17] jophish: that is from the root shell of the installer? [14:17] TJ-: yeah [14:17] jophish: have you (re)copied the iwlmvm.ko over there yet? [14:18] TJ-: yeah, It's in /lib/modules/$(uname -r)/kernel/drivers/net [14:18] jophish: what exact version info does "modinfo iwlmvm" report? [14:19] jophish: look at "srcversion" and "vermagic" and compare them to "modinfo iwlwifi" to ensure both modules match [14:21] TJ-: srcversion differs [14:22] jophish: yes, that's expected. Hmm, not sure what you have going on there then and I'm not in a position to test that here right now [14:22] I need to get on installing desktop now [14:22] thanks for the help though, TJ- [14:22] jophish: good luck with it [14:23] thanks :) [14:58] sat_: I only followed that guid to 50% for now, but I already think that the "cache" in your output is from the vgname (if you set up similar to that example you linked) [14:59] sat_: that could mean you might have to tune the other (caching) device instead of origin [14:59] sat_: I'll continue and let you know what I find [14:59] cpaelzer: thanks [15:00] cpaelzer: the key guide is at slide 22 [15:00] sat_: yeah already at step 7 or so there [15:00] sat_: found an nvme [15:06] zul, jamespage: ocata backports should be fixed up now [15:09] coreycb: k what did you do? [15:24] sat_: ok, I now understand it [15:24] sat_: it is actually written in the doc [15:24] sat_: see https://www.kernel.org/doc/Documentation/device-mapper/cache-policies.txt [15:25] sat_: TL;DR - the tunable you wanted to change is deprecated and war part of the "mq" policy which no more exists [15:25] sat_: It ended with "The following tunables are accepted, but have no effect:" [15:25] sat_: Stochastic multiqueue (smq) is the successor [15:26] sat_: and as one of the benefits to quote "smq also does not have any cumbersome tuning knobs." [15:26] cpaelzer: yes, I've seen that [15:26] sat_: so it is correct - there just are no tunables anmore with smq [15:26] sat_: you'd get the same message if you tried to set arbitrary key names [15:29] cpaelzer: that was my impression as well. I made the volumes with dmsetup directly and provided the arguments during construction, and still saw no performance difference [15:29] sat_: now that I have set it up I think I'm gonna do a basic perf eval and make an askubuntu post about it [15:30] I already have an equiv uncached refernce dev around and I'm used to fio [15:30] cpaelzer: take a look at this: https://www.redhat.com/archives/dm-devel/2016-April/msg00047.html [15:30] to help the next one wondering make something searchable by google then [15:30] * cpaelzer is looking [15:32] sat_: yeah I'm at kernel 4.8, so as I assumed mq = smq alias [15:35] cpaelzer: I was also getting smq by default on 4.4 [15:37] sat_: yeah default was changed longer in the past, but sicne 4.6 mq is only an alias for smq [15:37] sat_: and mq was dropped completely === unreal_ is now known as unreal [15:40] cpaelzer: that essentially means that performance is as is - no changes are possible with messages or constructor arguments [15:40] that's also useful info [15:47] zul, mostly fixed things in ca-patches for the backports [16:05] hello guys, i am not receiving mail at my local server, right now i am using a dynamic IP with a DNS name with reverse DNS. I have also check if this IP was not blacklisted and it is in one from a big list. It is possible that my ISP IP provider is blocking any SMTP connection from my external server to my local server? [16:06] I already did telnet to my external IP to port 25 and it was connecting [16:37] hi all, does anyone know how to deal with long ethernet device names and vlans? [17:01] Hello, I'm trying to install OpenStack using Conjoure-UP under a VMWare VM and the installation is failing, is the installation supported under a VM or should it be on bare metal? [17:01] dr4c4n: You want to change them? [17:14] hey guys, can someone help me with qemu tracing log backend? [17:21] Sorry ThiagoCMC I haven't touched it more than reading about it - so I'm only as smart as a search engine n this would be [17:22] ThiagoCMC: do you have a particular issue - might get more response if your share this [17:23] I have a VM that is doing a lot of vmexit, even if the guest is 99% idle, so, I need to figure out exactly what's happening here... [17:24] ThiagoCMC: which architecture? [17:24] amd64 [17:25] ThiagoCMC: perf kvm should be for that these days [17:25] Right! I'll try that... [17:25] gotta reboot, brb... thanks man! [17:25] not sure if all in the archive is enabled for it [17:25] but it shoudl be as it is aroudn for quite a while [17:25] sure, no problem... [17:25] ThiagoCMC: start here https://lwn.net/Articles/513317/ [17:25] but check for newer doc [17:25] ThiagoCMC: ./perf kvm stat report --event=vmexit [17:26] wow! [17:26] that's looks juice! [17:26] and that is 4 years ago [17:26] lol [17:26] nice [17:26] :-D [17:26] brb [18:00] hm, anyone know offhand how ansible interacts with /etc/modules? is there a "proper way" to ask ansible to setup a kernel module to be autoloaded on reboot? [18:01] TJ: sorry just got your message. I'm trying to add a vlan to an extremely long ethernet device name, but it says name too long as the error message when I try to add the vlan number tag to the end of the device name for ifconfig [18:08] coreycb:im going to hold off on uploading a newer brick [18:13] zul, ok np. as long as CI is successful we can try to batch up dependency updates. [18:43] Hello all, I'm trying to install OpenStack using Conjoure-up on a VMWare VM but the installation is continously failing, is the installation supported on a VM or I should install on a baremetal server? === JanC_ is now known as JanC [19:23] I'm struggling to get a fresh install to boot on an IBM System x3100 M4. Has someone achieved this? [19:27] guampa: i never tried, but according to https://lenovopress.com/tips0811-system-x3100-m4#supported-operating-systems it works with RHEL 7, so I guess it should work with at least Ubuntu 12.04 or 14.04 as well. [19:27] at which point do you get stuck? [19:38] I'm sorry, seems I got the channel wrong. It was for the Proxmox channel (which uses Debian anyway) [19:39] the system appears to boot or not boot after installed in a random fashion, or at the least I haven't been able to determine what BIOS setting is affecting the correct legacy boot [20:37] is anyone else having major slowdowns on ubuntu's repo lists? [20:38] i'm getting on average 25bps [20:38] Bps [20:39] change mirror perhaps? [20:39] any suggestions? [20:39] to a different mirror i mean [20:40] ghostmech007: which IP is slow for you? there may be something wrong with it.. [20:40] ghostmech007: mirrors.kernel.org has gobs of bandwidth, it usually works well for me [20:41] sarnold: i'm running the standard archive.ubuntu.com [20:43] ghostmech007: that's served by several machines; if you can spot which IP specifically is slow (via ss or netstat or lsof or fuser) that might be helpful [20:44] sarnold: i'll figure it out in a minute and get back to you [20:48] sarnold: from what I'm seeing it's the 91.189.88.162 machine [21:38] iirc broken/slow mirrors can be reported here: https://launchpad.net/ubuntu === skeezix-hf is now known as Ofir === Ofir is now known as skeezix-hf [22:53] I have this snippet https://gist.github.com/danielburrell/9a793380d2d7a4a9814f7febff859e93 i'm trying to migrate from upstart to systemd. I'm unsure about how I can migrate those pre-start directory creations, and more importantly the main exec command which does some fancy start-stop-daemon thing before executing the actual startscript... Anybody advise? [23:00] terabyte: this is a very handy cheat-sheet for converting https://wiki.ubuntu.com/SystemdForUpstartUsers [23:00] terabyte: you'll probably also need to spend an hour doing nothing but reading systemd docs, e.g. https://www.freedesktop.org/software/systemd/man/systemd.exec.html# [23:25] cpaelzer, hey man, that "perf trace" isn't detailed enough, for example: https://paste.ubuntu.com/23451853/ [23:25] It doesn't show, for example, which "kvm_msr: msr_write" register it is triggeting, is it 80b? No clue... === sikun is now known as sikun_