=== TheHonorableKitt is now known as THKitten [07:17] jamespage: what is the usual acceptance time for patches on OVS? [07:18] jamespage: I now have two acks and it is up quite some days - I just want to set my expectations right if we will get that into 2.11 as well or at least into master until we upload final 2.11 [07:25] on bionic, if i add -l to libvirtd_opts in /etc/default/libvirtd (as the file says to do), systemctl fails to restart/start libvirtd.service with 'respawning too fast'. should i file a bug? [07:29] i've had someone else confirm this, so it's not a local issue. [07:30] ducasse: let me take a look first [07:31] sure thing [07:32] ducasse: I'll test on disco, what release are you on so that I take a look there as well? [07:33] bionic [07:39] ducasse: once you enable listening on tcp you also need certificates - the issue after enabling -l is "Cannot read CA certificate '/etc/pki/CA/cacert.pem': No such file or directory" [07:40] you can use the package ssl-cert to create one or take a look at certbot/letsencrypt (for a real cert) I think [07:40] oh, i must have missed that when i looked at the log. perhaps there should be a note on this in /etc/default/libvirtd? [07:41] IMHO it is not really abug, if anything a note there would be nice [07:41] Good morning [07:41] yeah, i've already got a cert i can use [07:41] hi lordievader [07:41] morning lordievader [07:41] Hey cpaelzer , ducasse [07:41] How are you doing? [07:41] but thanks for taking a look, cpaelzer [07:42] well I'm not done [07:42] give it a try, but on bionic the error is less obvious - as you say when you checked the log [07:43] it is the same issue, even though the error looks slightly different and is not yet auto-marks as crit by journal [07:43] error : virNetTLSContextCheckCertFile:120 : Cannot read CA certificate '/etc/pki/CA/cacert.pem': No such file or directory [07:43] ducasse: so yes, using a cert shoul help you [07:45] Im having some problems with pptp-linux on 16.04. When using "poff", routes doesn't seems to reset correctly. Any idea how to fix this? I can always run "networking restart", but doesn't like it :P Route log: https://pastebin.com/eB4N3RV5 [07:45] ducasse: the config is in debian/libvirt-daemon-system.libvirtd.default (in packaging) and shared with Debian - I'd not want to add Delta jsut for that as we share the issue [07:46] ducasse: if you report a bug about adding a hint in the conffile (which is fine) would you report to Debian as well - when you open an ubuntu bug state the Debian bug number so we can track it from LP [07:46] ducasse: or to make things more easy you can only open a Debian bug and let me know the number [07:46] i could add a debian bug, ask for a brief note to be added. it wouldn't take many words [07:46] ducasse: excactly [07:47] ok, i'll do that later today, i'll update you with bug number [07:52] thanks ducasse [07:53] np, thanks for the help [07:54] btw man libvirtd will list all the certs/keys involved [07:55] so that you know all the paths at once [07:56] ducasse: for a qucik not recommended way out you could always set listen_tls = 0 listen_tcp = 1 in /etc/libvirt/libvirtd.conf [07:57] that will make tcp listening "work" but well without security it isn't as nice :-) [07:57] just saying, in case someone reads this and just wants it to listen on the network without the need for certs [08:00] yeah, i think i'll use a cert (even though there's nothing to target on this system) [08:11] maybe someone is familiar with intel_pstate and would know why ubuntu server instalation could not use it by default for xenon E5 processor? === lotuspsychje_ is now known as lotuspsychje [09:59] cpaelzer: I dug into the ovs build failure and can reproduce it against the bionic/stein UCA [09:59] the -I/usr/include/dpdk gets a trailing , so none of the dpdk related imports work [10:02] a trailing colon - hmm [10:02] why did that work for me [10:02] jamespage: can you log onto any of those builds and check what pkg-config returns [10:03] to see if it is in broken in the pkg-config of dpdk or somewhere else [10:03] cpaelzer: pkg-config return looks fine [10:03] hmm [10:03] ok [10:03] one sec - its just running tests... [10:03] jamespage: so what yould be the difference to my build that I linked a few days ago? [10:03] y->c [10:03] "-include rte_config.h -march=corei7 -I/usr/include/dpdk/../x86_64-linux-gnu/dpdk -I/usr/include/dpdk" [10:04] the march is a ugly beast, but I discussed it with many people there seems to be no way around - I don't see your colon thou [10:05] cpaelzer: I think if I tweak one of the patches like this - http://paste.ubuntu.com/p/QhNtDJWQqp/ [10:05] it resolves the issue [10:06] jamespage: that seems to make no sense to me yet [10:06] that colon is a colon for the autoconf handling [10:06] and it seperates differnt actions [10:07] one is the assign of PDDK_INCLUDE and the second one is the assign of DPDK_LIB [10:08] cpaelzer: comma needs to be outside of [] ? [10:08] no [10:08] https://www.irccloud.com/pastebin/KSuoLJlt/ [10:08] this is the definition PKG_CHECK_MODULES(prefix, list-of-modules, action-if-found, action-if-not-found) [10:08] and in our case we are reaching the path of action-if-found [10:09] which we set as [10:09] [DPDK_INCLUDE="$DPDK_CFLAGS", DPDK_LIB="$DPDK_LIBS"] [10:09] the colon inside of this separates multiple elements of that action [10:09] at least I thought so [10:09] that evaluates to [10:10] DPDK_INCLUDE="-I/usr/local/include/dpdk -I/usr/include/dpdk", DPDK_LIB="-ldpdk" [10:10] in the script [10:10] that is wrong, so much I agree [10:10] but it should be two separate actions [10:10] and btw - if you are in the path that you mentioned then you are wrong [10:10] because that is the "no pkg config found path" [10:10] you should be hitting the one above it [10:11] the one I listed 12 lines above here [10:11] yes but even that will append the , [10:11] [DPDK_INCLUDE="$DPDK_CFLAGS", DPDK_LIB="$DPDK_LIBS"] [10:11] yes that one should be yours [10:15] https://stackoverflow.com/questions/12735432/how-to-pack-multiple-statements says newlines instead of commas [10:15] arr I'd like to find the right official definition to be sure [10:15] https://www.irccloud.com/pastebin/vyQ1pt9I/ [10:16] well that matches what I had above [10:16] cpaelzer: re-confirmed - the patch as is generates [10:16] DPDK_INCLUDE="$DPDK_CFLAGS", DPDK_LIB="$DPDK_LIBS" [10:16] and then the checks for features fail as we get the trailing comma [10:16] I mean the definition for "one" of those [10:16] yeah that would be wrong for sure then [10:16] [10:18] maybe more reasbale with newlines [10:18] let me try [10:18] trying that now [10:19] jamespage: http://paste.ubuntu.com/p/52S6ftNZ2g/ [10:19] would that work for you [10:22] it appears to be [10:24] I still don't get why it worked for me and the OVS travis Ci then [10:24] I have prepared a mail to the discussion of the patch upstream [10:24] jamespage: please give me a ping once you can confirm that this resolves your build [10:24] I'll hit send then [10:25] I should see the result of that in my build log right ... [10:30] jamespage: I mean on the disco build of the same I literally run traffic over OVS-dpdk ports so that must have been working [10:31] cpaelzer: its certainly odd [10:31] jamespage: here a gcc line of my 18.04 backport PPA [10:31] building OVS with dpdk enabled against 18.11 [10:31] gcc -DHAVE_CONFIG_H -I. -I.. -I ../include -I ./include -I ../lib -I ./lib -Wdate-time -D_FORTIFY_SOURCE=2 -Wstrict-prototypes -Wall -Wextra -Wno-sign-compare -Wpointer-arith -Wformat -Wformat-security -Wswitch-enum -Wunused-parameter -Wbad-function-cast -Wcast-align -Wstrict-prototypes -Wold-style-definition -Wmissing-prototypes -Wmissing-field-initializers -fno-strict-aliasing -Wswitch-bool -Wlogical-not-parentheses -Wsizeof-array- [10:31] argument -Wbool-compare -Wshift-negative-value -Wduplicated-cond -Wshadow -mssse3 -include rte_config.h -march=corei7 -I/usr/include/dpdk/../x86_64-linux-gnu/dpdk -I/usr/include/dpdk -I/usr/include/libnl3, -D_FILE_OFFSET_BITS=64 -g -O2 "-fdebug-prefix-map=/<>=." -fstack-protector-strong -Wformat -Werror=format-security -MT lib/odp-util.lo -MD -MP -MF lib/.deps/odp-util.Tpo -c ../lib/odp-util.c -o lib/odp-util.o [10:31] could you check your build for the same line and post it [10:32] I'd be eager to see the difference [10:32] as there would be the colon that you assume is rendered into that string right? [10:32] well you might not get that far as your configure breaks already :-/ [10:32] jamespage: ^^ [10:33] cpaelzer: odd - you still have a trailing , - its just after libnl3? [10:35] hmm yeah - thanks for spotting [10:35] but it is not breaking it oO [10:35] jamespage: once your build is complete using the newlines please let me know how this line looks like then [10:36] cpaelzer: yep doing that now - testing bionic and disco [10:55] cpaelzer: https://launchpad.net/~james-page/+archive/ubuntu/stein LGTM [11:00] I must admit I don't understand it [11:00] but let me send the mail to ovs-dev [11:02] sent, you are on CC in cae there are specific questions about you encountering the issue at first [11:03] I think so [11:03] oh I am yes!" [11:03] doh, [11:03] is that the trailing comma [11:03] after doh [11:04] 😜 [11:38] jamespage: we have an ack and prior art on the newline suggestion [11:38] and bluca is really good at autoconf [11:38] I'll respin my patch for OVS-dev [11:45] great [11:46] jamespage: sent [12:36] Hi, can anyone help me with an issue doing a UEFI install on a Hyper-V gen 2 VM? [12:36] I've got a preseed working fine for Bionic, I'm trying to do one for Xenial [12:36] The install completes and the machine shuts down [12:36] but then I can't boot into it, unless I go to recovery, type exit and it puts me into it properly [12:36] Preseed, images, explanation I've wrote up here: https://pastebin.com/raw/YbRD73j4 [12:44] why do you need/want UEFI in a VM? [12:46] I don't need/want it, but with Gen 2 you get access to better devices [12:47] It works for Bionic, just not Xenial [12:48] not sure how to troubleshoot this any further [12:51] also leftyfb, this is in test for rolling out to our users, that's the primary reason for running it in Hyper-V [12:52] It will be going to bare metal when it works [12:53] So I'd like to have UEFI/secure boot working for that [12:55] Hi, I've got a Linode VPS on which I'm running Ubunutu. I run a web server, but I handle mail on a separate shared hosting account. I have a wordpress plugin that I'd like to be able to send mail. I found this guide https://www.linode.com/docs/email/postfix/postfix-smtp-debian7/ is that the best way to set things up? [12:59] boblamont: you don't need a full MTA if all you want is to use an external one. take a look at dma (Dragonfly Mail Agent) [12:59] !info dma [12:59] dma (source: dma): lightweight mail transport agent. In component universe, is optional. Version 0.11-1build1 (bionic), package size 48 kB, installed size 148 kB [12:59] there's also ssmtp, nullmailer and a couple others, i think [12:59] boblamont: also note that wordpress can probably be configured to use an external MTA directly via some smtp lib [13:00] ssmtp is imho abandonware at this point [13:00] thanks, I wondered if that was possible... I want to install and configure as little as possible since it's just to send emails out of a single plugin [13:01] many php web apps can do smtp and smtp-auth, and hopefully also tls nowadays. [13:02] personally I always prefer a locally available mail transport so it can also send mail for cron etc, or locally deliver mail to root (eg. some problem reports). but if you don't need/want that, then definitely use a smtp lib based WP plugin to talk to the external MTA directly [13:21] jamespage: well i think we're mostly narrowed down to neutron-ish backport failures for the stein cloud archive. i'll handle the vitrageclient one [13:21] mostly just waiting on os-key i think for neutron [13:31] that was easy, a plugin, a wizard and a successfully received test email, thanks for pointing me in the right direction! [13:41] coreycb: hello o/ are you around? [13:42] i took some time to struggle trough ubuntu 18.04 + py3 + stein in puppet openstack just one thing left [13:42] http://paste.openstack.org/show/744673/ [13:42] python3-websockify is not installed for nova-novncproxy [14:54] tobias-urdin: we've historically had issues with getting websockify into main due to security concerns so we can't add it to nova Depends. we should have it in nova Suggests though and that is a bug. [14:56] tobias-urdin: oh maybe we're good to go now. bug 1108935. jamespage do you agree we can add websockify to nova Depends now? [14:56] bug 1108935 in spice-html5 (Ubuntu) "[MIR] websockify, spice-html5" [High,Confirmed] https://launchpad.net/bugs/1108935 [14:59] coreycb: +1 [15:00] coreycb: tbh that should have been on the depends anyway, but just unseeded at the nova-novncproxy level [15:00] jamespage: ok cool. i'll add it. [15:01] tobias-urdin: i'm adding that dependency and you'll see it soon. thanks as always for the feedback. :) [15:33] Hi guys, my routes are not added using netplan... any idea ? [15:33] lotuspsychje ok :P [15:34] gislaved, can you pastebin your netplan config? [15:34] lordcirth__ not really as I'm on KVM but I use the netplan examples [15:34] without seeing your config it's hard to give guidance [15:35] gislaved, cat /etc/netplan/* | nc termbin.com 9999 [15:35] The link should be short enough to retype [15:35] I know but it's more like when I get the same info from DHCP and set it staticly should that work ? it's a VPS and it doesn't surprise me as they don't allow static IP set [15:36] lordcirth__ how would that be possible when I don't have a route ;) [15:36] gislaved: i'd use DHCP at least to get a route in the interim [15:36] then share the configs you're using. It's possible the routes you think are being added *aren't* [15:36] DHCP does some weird things like that :P [15:37] yes it does as it's a /32 ip and the GW is a /24 lower :S [15:37] on dhcp [15:37] gislaved: then your netmask needs to be /24 [15:37] not /32 [15:37] coreycb: thanks :) [15:37] for your setup. it won't allocate the entire range but it DOES state how to actually hand things around [15:37] A /32 IP cant reach anything outside it, the netmask is actually /24 typically [15:38] tewaroh man that I ddin't try that [15:38] or /26 or /18 or {INSERT CIDR RANGE HERE} depending on how they split things around [15:38] no but why does DHCP the /32 ? [15:38] gislaved: i had the same headaches at times setting IPv6 up on a VPS, where the GW was in the /48 above a /64 [15:38] as it's single assigned ? [15:38] gislaved: no idea, but if you're static-ing it use /24 [15:38] teward ok I feel less stipud :P [15:38] gislaved: To be fair [15:38] stupid, I hate these VPS providers [15:38] I don't know [15:38] but the consideration point is this: [15:39] tell me [15:39] I have a VPS assigned 1.2.3.4. The gateway is 1.2.3.0. The subnet for the IP address on the *server* may be a /32 single assigned IP address, but the gateway sits in the /24 subnet, so you have to adjust the subnet to be a /24 on the IP itself [15:39] this is just how 'subnetting' works [15:39] now it still doesn't apply my nameservers, I need to find out why [15:39] yap [15:39] gislaved: you should probably focus on *getting the thing back on the net* first :P [15:40] it's on the net [15:40] oh no not yet [15:41] sorry my misstake :) I turned on dhcp but on a /24 my route is there [15:43] well again as i said i'd need to see the configs [15:43] even if you just take a picture of what you see on your screen [15:43] and upload that picture to an image hosting site :P [15:44] the problem is without *seeing* your configs it's impossible to know what's going on [15:44] and I'm sure lordcirth__ and others will say the same :p [15:45] I need to know the subnet of the vps [15:45] and gw [15:45] gw is what I know but still [15:45] you can probably get the subnet from `ip -4 addr list` after DHCP [15:45] but i'd probably reboot the VPS to make it 'refresh' its network configs cleanly [15:46] so we don't have any cruft lying around in there [15:46] and stop trying to apply your stuff temporarily [15:46] start with "working" then we'll get details. [15:47] nah /32 for dhcp [15:51] i can still figure out the stuff from once it's working, because I know things [15:51] I KNOW THINGS! [15:51] but still [15:51] COFFEE TIME, back in a bit [15:51] rbasak: may I add to the trello board adding the 3rd party plugin to the nginx stuff? As a to-do item for me because GeoIP is now paid-only [15:51] (because of MaxMind) [15:51] rbasak: it's got some funky chaos in there for building modules that I have to debug first, so I need a reminder item :P [15:51] coreycb: do you know if swift packages is py3 yet? seems like it's running with py2.7 [15:54] (did it anyways rbasak xD) [15:54] * genii 's ears perk up momentarily at the mention of coffee [15:54] tobias-urdin: it's py2. upstream is py2 only so far. [15:54] unfortunately [15:55] coreycb: ImportError: No module named keystonemiddleware.auth_token [15:55] maybe it doesn't depend on the keystonemiddleware py2 pkg [15:56] teward: sure, no problem [15:57] rbasak: cool. just as an FYI, if Upstream doesn't move quickly to geoip2 it may end up with a MIR because of GeoIP being pretty important :P [15:57] though come that time we'll have to visit whether we want to adjust -core to actually include the module [15:57] because as I said, GeoIP is a pretty useful function in the core modules :| [15:57] but that's down the road yet :P [15:59] coreycb: another question, nova packages has never depended explicitly on python-rbd/python-rados/python-ceph right? [16:00] mwhudson: just fwi, i am really greatful for 'go' snap. [16:05] rbasak: saw your response to the backports proposal, i'll review when I get a minute :P [16:06] probably poke Laney too since Laney made a bunch of the suggestions there : [16:06] P [16:06] bleh stupid keyboard again [16:10] teward: need SRU information in bug 1808882 prior to the upload of the SRU debdiffs please. [16:10] bug 1808882 in chkrootkit (Ubuntu Disco) "false positive on tcpd" [Low,Triaged] https://launchpad.net/bugs/1808882 [16:10] rbasak: 1 moment [16:10] The Disco upload looks good to go [16:10] ERR:SLOW [16:10] maybe IS should prod the wiki server to speed it up :| [16:15] rbasak: added though I'm bouncing around between 10 tasks [16:15] so let me know if i missed anything [16:16] tobias-urdin: interesting question re: ceph deps. i don't think they ever have had explict deps on them. we generally only depend on what's in (test-)requirements.txt [16:17] Sure, thanks. [16:18] we had a major AD outage that fubar'd all email on the work network so I'm helping to pick up the pieces and make things work again so :| [16:18] (hence 10 tasks) [16:18] tobias-urdin: did you hit the swift keystonemiddleware error on install? [16:21] Hi, can anyone help me with an issue doing a UEFI install on a Hyper-V gen 2 VM? I've got a preseed working fine for Bionic, I'm trying to do one for Xenial. The install completes and the machine shuts down but then I can't boot into it, unless I go to recovery, type exit and it puts me into it properly [16:21] Preseed, images, explanation I've wrote up here: https://pastebin.com/raw/YbRD73j4 [16:21] I've tried the HWE and GA kernel [16:23] coreycb: swift-proxy does not start because python-keystonemiddleware is not installed, i assume it has worked before because some other package has always installed keystonemiddleware py2 [16:23] but now only swift is py2 i think swift itself should depend on that [16:24] tobias-urdin: probably so, have a traceback by any chance? it might be that another pkg needs it b/c i'm not seeing an import other than in tests for swift. [16:25] tobias-urdin: maybe keystone... this is tricky [16:25] coreycb: http://logs.openstack.org/13/635513/1/check/puppet-openstack-integration-5-scenario002-tempest-ubuntu-bionic/2a97889/logs/swift/proxy.txt.gz [16:27] swift can be used without keystone, but with keystone middleware loaded by paste it will fail [16:53] tobias-urdin: thanks. i'll dig in more later but for now we have bug 1815093. [16:53] bug 1815093 in swift (Ubuntu) "swift-proxy fails to start on stein - missing python-keystonemiddleware" [Undecided,New] https://launchpad.net/bugs/1815093 [17:11] teward: mind if I change your bionic SRU version string to 0.52-1ubuntu0.1? 0.52-1ubuntu1 won't cause a technical problem but I prefer making it obvious it's an SRU by using the security team's scheme (which I think we should make more official for SRUs anyway) [17:13] teward: https://launchpadlibrarian.net/408977730/lp1808882-cosmic.debdiff has a spurious duplicate patch I think? [17:20] rbasak: wrt string: go ahead [17:21] rbasak: yep looks like quilt exploded. i can fix that in a bit. [17:21] that's not a hard fix :P [17:22] rbasak: cosmic/disco had additional patches past 23 which is why it fubar'd did Disco have the same issue? [17:24] rbasak: https://launchpadlibrarian.net/410209868/lp1808882-cosmic.debdiff <-- this should be much cleaner [17:24] for Cosmic [17:24] OK thanks! [17:25] rbasak: and regarding version strings, if it's off minorly like that then sure go ahead [17:25] sometimes reading through the sec team's version schemes is tricky :P [17:25] rbasak: actually [17:25] rbasak: if this document is wrong, then you need Security to update it [17:25] https://wiki.ubuntu.com/SecurityTeam/UpdatePreparation#Update_the_packaging shows: [17:26] ''Previous version'' ''Security update'' [17:26] 2.0-2 2.0-2ubuntu0.1 [17:26] oops multiline paste :| [17:26] but that's why 0.1 was used [17:26] rbasak: if *this* is in difference with the sec team version policy then the policy needs updated [17:26] cc sarnold ^ because sec team [17:26] I think that document is correct. [17:27] Previous version was 0.52-1 in one release only. [17:27] So what would you use? [17:27] oh WOW i'm blind [17:27] rbasak: E:NOCOFFEE [17:27] rbasak: on my side it's 0.1 that patch as 1 [17:27] No worries :) [17:27] so maybe I uploaded the patch, fixed it, then forgot to reupload :| [17:27] *derps* [17:27] ten bucks says dch did it the first time >.> [17:28] rbasak: i also misread what yo uwere saying, my bad. [17:28] that's the OTHER problem that lack of sleep causes >.> [17:29] rbasak: my apologies for not paying attention enough :| [17:29] thankfully i have a day off tomorrow so :P [17:30] rbasak: thanks for catching the version string issue as well, that's actually the most MINOR of the issues heh [17:43] rbasak: thanks for sponsoring :) [17:44] stupid IRC client >.. [18:04] rbasak: did SRU get subscribed to that bug? [18:04] or do we need to add them [18:37] teward: no subscription needed. They'll pick it up from the queue. [18:37] ack [18:46] smoser: np, i keep meaning to have another run at getting go upstream to maintain it [18:46] (why do they keep doing releases on fridays dammit) [19:08] rbasak: looks like it's landed now for Disco, thanks :) [19:18] teward it's an ubuntu issue I have [20:22] do I need to worry about this? "device-mapper: reload ioctl on osprober-linux-sdb1 failed: Device or resource busy" -- I got a bunch of them while updating: http://paste.ubuntu.com/p/hS7YTdkRZk/ [20:23] no idea [20:23] "Found Ubuntu 11.04 (11.04) on /dev/sdn3" sweet :) [20:24] os-prober is the 2nd package I purge on a machine, the first being nano ;) [20:24] sdeziel: oh yeah?? I friggen hate those messages in my dmesg.. [20:24] sdeziel: please promise me it's okay to purge? :) [20:25] sarnold: IIRC, it probes every block devices looking for bootable OS/bootloaders, not something I need nor want [20:25] that 11.04 image.. I popped the drive in to try to find a lost file on an old source control snapshot (it wasn't there) ... and left it in hoping that it would encourage me to actually copy off the data that I want :D [20:25] sarnold: I especially don't like when my hypervisor tries to probe the disks assigned to my (running!) VMs [20:26] sdeziel: and new grub packages, new linux-image packages, those install and work fine afterwards? [20:26] sarnold: oh yeah [20:26] sdeziel: thanks! purging.. :) [20:26] If you aren't dual-booting, you don't need os-prober [20:28] for sarnold, it's probably more octo-booting though ;) [20:29] I don't even like single-booting [20:29] I think my record was 7 [20:29] I want to pretend my machine is always running [20:29] back when I did dual-boot, I did it by yanking one drive and inserting another. to make sure nothing got bright ideas about scribbling over the other os [20:29] kexec'ing another OS must be fun [20:29] sdeziel: WAY FASTER reboots [20:35] kexec is the bestest. [20:37] rebooting 2 socket servers with like 4-5 seconds of downtime. [20:48] exactly, no more five minutes for who knows what it's doing :) [20:50] Although, actually spinning down hard drives occasionally is a good idea - you don't want to find out they don't spin up all at the same time. [20:50] Assuming you actually have spinning disks. [20:58] indeed this machine's got a bunch of em