Saturday, 2019-10-05

openstackgerritMerged opendev/system-config master: Install jeepyb into the gerrit images  https://review.opendev.org/68260100:07
*** factor has quit IRC00:09
*** yamamoto has joined #openstack-infra00:10
*** gyee has quit IRC00:15
openstackgerritMerged openstack/diskimage-builder master: Add environment switch for centos8 to use dnf  https://review.opendev.org/68445900:18
corvusyay the password is in place; i'll restart the executors now00:29
fungilmk if you need help with it00:31
fungithis fudge cools at a rate of ~1c/min00:31
corvusfungi: would a fan help?00:32
corvus(do you want fudge to cool slowly or quickly?)00:32
fungimaybe, but doing my best not to speed it up, yeah00:32
fungislowly00:32
corvusexecutors cool slowly too00:33
fungiwell, at least per the 1963 good housekeeping recipe. the 1938 fannie farmer recipe calls for working it on a marble slab, i should have one but do not at present00:33
fungithis is very reminiscent of tempering chocolate00:34
corvusall stopped00:37
corvusstarting up again00:38
corvusand i've rechecked 678875 to exercise it00:38
fungigrafana says we're back to 1200:39
fungiand 20 mergers00:40
openstackgerritJames E. Blair proposed zuul/zuul master: Get ansible version at start of job  https://review.opendev.org/68685300:46
corvussaw that in the logs when it shut down ^00:46
corvusotherwise things lgtm00:47
fungiyeah, that looks like a good idea00:49
*** ricolin_ has joined #openstack-infra01:03
*** yamamoto has quit IRC01:04
*** yamamoto has joined #openstack-infra01:08
*** slaweq has joined #openstack-infra01:11
*** slaweq has quit IRC01:16
*** yamamoto has quit IRC01:21
*** michael-beaver has quit IRC01:33
*** yamamoto has joined #openstack-infra02:38
*** yamamoto has quit IRC02:48
*** yamamoto has joined #openstack-infra02:59
openstackgerritMerged openstack/diskimage-builder master: Use $YUM instead of direct calls in more places  https://review.opendev.org/68446002:59
openstackgerritMerged openstack/diskimage-builder master: yum-minimal: Don't install yum, install libcurl  https://review.opendev.org/68446102:59
*** whoami-rajat has joined #openstack-infra03:12
*** yamamoto has quit IRC03:43
*** yamamoto has joined #openstack-infra03:48
*** factor has joined #openstack-infra03:53
*** yamamoto has quit IRC04:04
*** slaweq has joined #openstack-infra04:11
*** slaweq has quit IRC04:16
*** dklyle has quit IRC04:36
*** david-lyle has joined #openstack-infra04:36
*** yamamoto has joined #openstack-infra04:36
*** yamamoto has quit IRC04:47
*** david-lyle has quit IRC04:58
*** dklyle has joined #openstack-infra04:59
*** ociuhandu has joined #openstack-infra05:01
*** cyberpear has quit IRC05:04
*** ociuhandu has quit IRC05:05
*** yamamoto has joined #openstack-infra05:22
*** dklyle has quit IRC05:25
*** whoami-rajat has quit IRC05:31
openstackgerritMerged openstack/diskimage-builder master: pip-and-virtualenv : deprecate source for CentOS 8, new variables  https://review.opendev.org/68446205:53
*** kjackal has joined #openstack-infra06:01
*** slaweq has joined #openstack-infra06:11
*** slaweq has quit IRC06:16
*** e0ne has joined #openstack-infra06:16
*** e0ne has quit IRC06:18
*** jbadiapa has quit IRC06:22
*** yamamoto has quit IRC06:35
*** yamamoto has joined #openstack-infra06:40
*** yamamoto has quit IRC06:40
*** yamamoto has joined #openstack-infra06:40
*** yamamoto has quit IRC06:44
*** whoami-rajat has joined #openstack-infra06:46
*** dpawlik has joined #openstack-infra07:03
*** slaweq has joined #openstack-infra07:11
*** kjackal has quit IRC07:13
*** slaweq has quit IRC07:15
*** slaweq has joined #openstack-infra07:18
*** yamamoto has joined #openstack-infra07:20
openstackgerritMonty Taylor proposed opendev/system-config master: Remove dependencies from promote  https://review.opendev.org/68686807:20
mordredcorvus: https://zuul.opendev.org/t/zuul/build/0462c9784eb84765a83e346a3f0d7cd4 <-- upload image error with an http issue07:23
*** yamamoto has quit IRC07:27
*** yamamoto has joined #openstack-infra07:50
*** prometheanfire has quit IRC07:53
*** prometheanfire has joined #openstack-infra07:54
*** yamamoto has quit IRC07:54
*** xenos76 has joined #openstack-infra08:17
*** xenos76 has quit IRC08:32
*** ociuhandu has joined #openstack-infra08:52
*** ociuhandu has quit IRC08:54
*** markvoelker has joined #openstack-infra09:00
*** markvoelker has quit IRC09:05
*** e0ne has joined #openstack-infra09:42
*** yamamoto has joined #openstack-infra09:53
*** xenos76 has joined #openstack-infra09:55
*** jistr has quit IRC10:23
*** jistr has joined #openstack-infra10:23
*** xenos76 has quit IRC10:31
*** e0ne has quit IRC10:36
*** yamamoto has quit IRC10:47
*** ociuhandu has joined #openstack-infra10:55
*** dpawlik has quit IRC10:57
*** ociuhandu has quit IRC11:00
*** xenos76 has joined #openstack-infra11:06
*** e0ne has joined #openstack-infra11:18
*** yamamoto has joined #openstack-infra11:24
*** markvoelker has joined #openstack-infra11:29
*** markvoelker has quit IRC11:33
*** yamamoto has quit IRC11:35
*** yamamoto has joined #openstack-infra11:44
*** rh-jelabarre has quit IRC11:48
*** odicha has joined #openstack-infra11:50
*** yamamoto has quit IRC12:04
*** e0ne has quit IRC12:04
*** factor has quit IRC12:07
*** factor has joined #openstack-infra12:07
*** yamamoto has joined #openstack-infra12:09
*** yamamoto has quit IRC12:13
*** tosky has joined #openstack-infra12:20
*** exsdev has quit IRC12:52
*** e0ne has joined #openstack-infra12:54
*** yamamoto has joined #openstack-infra12:58
*** whoami-rajat has quit IRC13:01
*** xenos76 has quit IRC13:02
*** dave-mccowan has joined #openstack-infra13:08
*** yamamoto has quit IRC13:09
fungiyeah, 686853 is consistently hitting a POST_FAILURE on zuul-upload-image13:14
donnydis there an issue with nodepool being able to reach FN13:16
fungihttps://zuul.opendev.org/t/zuul/builds?job_name=zuul-upload-image says it's the only change to run that job since the registry switch13:16
donnydIt could be grafana just being a little behind, but it says its deleteing 102 test nodes and they have been gone for quite a while13:16
fungidonnyd: i don't think it's connectivity. the only fn errors which jump out at me in the nl02 logs are about insufficient quota13:31
fungiNot enough quota remaining to satisfy request 200-000561169913:32
fungioh, here we go...13:33
funginodepool.exceptions.ServerDeleteException: Timeout waiting for server 6cdf39ad-7748-466b-8d07-5665af1e08d7 deletion13:33
fungiso maybe there are deletion problems?13:33
fungilooks like it started trying to delete that one at 12:57:0013:34
fungi2019-10-05 12:57:00,573 INFO nodepool.DeletedNodeWorker: Deleting used instance 6cdf39ad-7748-466b-8d07-5665af1e08d7 from fortnebula-regionone13:34
donnydOh those servers are deleted13:41
donnydthey have been for an hour13:41
donnydOk, looks like its running fine again13:41
donnydweird, because it also looked like it happened  a couple times http://grafana.openstack.org/d/3Bwpi5SZk/nodepool-fortnebula?orgId=1&from=1570270402968&to=157028264008913:42
donnydoh well... back to normal now13:42
donnydmy DC is loving this cooler weather.. down to 72 degrees in there13:43
*** e0ne has quit IRC13:47
fungiyeah, we were finally able to turn off our air conditioning and open our windows again last night for the first time in ages13:56
*** Lucas_Gray has joined #openstack-infra14:01
*** goldyfruit_ has quit IRC14:06
fungiso back to the zuul-upload-image failures, both seem to have broken at the same spot, in the "Push tag to intermediate registry" task, the skopeo command died with "Error writing blob: Error determining upload URL: http: no Location header in response"14:18
AJaegerskopeo? See https://review.opendev.org/68680314:19
AJaegerfungi: do we need this one? ^ didn't review, just remembered seeing it flying by14:20
*** whoami-rajat has joined #openstack-infra14:24
fungioh, quite possibly. it doesn't mention that error exactly, but makes sense that it could be related14:24
AJaegertristanC: do we need that change? See 686853 and discussion above14:25
*** kambiz has joined #openstack-infra14:28
tristanCAJaeger: 686803 doesn't fix a missing Location header in response... so did skopeo got updated recently?14:32
fungiOct  5 01:46:32 insecure-ci-registry01 bdff2d543272[23452]: INFO:registry.api:Start upload POST zuul/zuul-executor uuid 25557bf2b1fc49a79115a2567af5b1fe digest None14:32
tristanCis the zuul-registry already in use?14:33
fungithat's the log from the zuul-registry service corresponding to this i think: https://zuul.opendev.org/t/zuul/build/0462c9784eb84765a83e346a3f0d7cd4/log/job-output.txt#1106014:34
*** xenos76 has joined #openstack-infra14:34
fungitristanC: yeah, as of ~18:00z yesterday14:35
tristanCi've rechecked 686803 to see if it reproduces14:36
*** pcaruana has joined #openstack-infra14:36
fungiit did it a couple times already, but can't hurt to collect more examples14:37
corvusi'm looking at the logs for the registry and seeing if we need more logging14:37
tristanCperhaps this is because of https://github.com/containers/skopeo/commit/7922028d7c994c5a489283dbf04f32a7dca44f5e14:38
tristanCwhich touch a file that has the "Error determining upload URL" message14:39
corvustristanC: did your local test with skopeo use that version?14:39
tristanCcorvus: no, i use the fedora package which only provides released version.14:39
tristanCiirc the ppa is built continously and it may have picked that commit14:39
corvusok, i'll grab skopeo from the ppa and test that locally14:40
fungiseems like that ppa for bionic hasn't been updated for weeks14:41
*** lxkong has quit IRC14:41
tristanCcorvus: seems like the POST now excepts a location header14:42
fungiskopeo 0.1.39-1~ubuntu18.04~ppa114:42
tristanCthe location should point at the PUT endpoint14:43
corvustristanC: it sounds like you've found the error, but i still don't see it and can not reproduce; tell me more14:47
corvuslocally, i've installed skopeo, ran into the bug that your skopeo patch fixed, i applied your patch locally, and a skopeo copy from a file to the registry works.  i haven't seen the Location header error14:47
tristanCcorvus: don't know yet, i'm currently building latest skopeo to try to reproduce locally14:47
tristanChum, with the master tip of skopeo?14:48
corvusfrom the projectatomic ppa14:48
corvusi will try again after building a zuul image -- maybe that will reproduce14:49
fungitristanC: https://launchpad.net/~projectatomic/+archive/ubuntu/ppa14:49
corvus(specifically, a zuul-executor image)14:50
fungitristanC: skopeo 0.1.39-1~ubuntu18.04~ppa1 from there is what it looks like we're using14:50
fungibuilt a couple months ago, 2019-08-1014:51
tristanCoh, and does that job ever worked with this skopeo and the zuul-registry?14:52
openstackgerritJames E. Blair proposed zuul/zuul-registry master: Add support for skopeo copy  https://review.opendev.org/68680314:52
openstackgerritJames E. Blair proposed zuul/zuul-registry master: Add timestamps to logs  https://review.opendev.org/68689514:53
fungitristanC: no, not in the two times it's run since switching to zuul-registry, that i can tell14:53
corvusgah, i didn't mean to rebase tristanC's patch14:53
corvuslooks like it's a no-change rebase14:53
corvussorry about that14:53
corvustristanC, fungi: can we go ahead and merge those 2 changes?  ^14:54
fungiyeah, already reviewing14:54
corvussince we don't have the access logs, it's possible we didn't log whatever is tripping up skopeo14:54
openstackgerritJames E. Blair proposed opendev/system-config master: Set debug logging in intermediate registry  https://review.opendev.org/68689614:56
corvusfungi, mordred: also ^14:56
fungicorvus: as for timestamps, i correlated based on what got copied into syslog since there's already timestamping. are these other logs, or just timestamping them when recording to a different place?14:57
corvusfungi: oh neat i didn't think to check that :)14:58
corvusfungi: i was just looking at "docker logs"14:58
corvusfungi: should we omit timestamps then?14:58
fungii was having trouble figuring out where to find the docker config file to make `docker-compose logs` work, then noticed docker was set to log via journald which is in turn copying to syslog14:58
corvusfungi: "docker logs" has a "-t" option which means that both ways we collect logs optionally have timestamps14:59
fungii'm fine with timestamps appearing in docker logs output too14:59
fungiahh14:59
corvusi'll rework that to omit timestamps for now, since it might be duplicative14:59
fungiwfm14:59
corvusfungi: the cherrypy access logs have an internal timestamp too (since they are standard web access log format) so that would make 3 if you viewed it in syslog.  that seems gratuitous15:00
fungiyo dawg i heard you like timestamps so...15:00
fungialso interesting that we're getting urllib3 debug logging already in syslog. is that being set elsewhere?15:01
openstackgerritJames E. Blair proposed zuul/zuul-registry master: Add debug env variable  https://review.opendev.org/68689515:01
fungiahh, i bet it's logging from openstacksdk15:01
fungibecause i see swift client interactions being logged too15:02
corvusfungi: yeah, but the logger is supposed to be at info.  i'm confused by that.15:02
corvusthat's not what happens for me locally, but i'm not running it in a container15:02
corvuswe're basically already getting debug level logging, except we're not getting the access logs, which i think we need for this.  also, arguably, we should move the access logs to the standard info level anyway, but later.15:03
fungiyeah, i think early on having debug logs is a good idea15:04
fungiwhen we decide we're ready to turn them off we can figure out why we're still getting debug logging from openstacksdk15:05
corvusi reproduced15:05
fungithe skopeo error?15:06
corvusyeah, the one we saw in prod: http://paste.openstack.org/show/781181/15:06
corvusi have no idea why that didn't show up in the container logs15:06
funginuts15:07
fungistderr instead of stdout maybe?15:07
openstackgerritMerged zuul/zuul-registry master: Add support for skopeo copy  https://review.opendev.org/68680315:09
fungiso the request header is too large for cherrypy? what are we stuffing in there which would bloat it?15:09
corvusi don't think it's headers15:10
corvusi think it's something about cherrypy and upload sizes15:13
corvusit may have a limit15:13
fungiahh, yeah it's accumulating the bytes read and if the next chunk would put it over maxlen it raises15:14
*** jamesmcarthur has joined #openstack-infra15:14
*** jamesmcarthur has quit IRC15:14
fungithe ChunkedRFile class seems to accept maxlen as an argument so maybe it's tunable15:15
*** jamesmcarthur has joined #openstack-infra15:15
tristanCcorvus: i was able to reproduce locally using a sufficiently large image, the zuul-registry yield OSError: Request Entity Too Large15:17
*** jamesmcarthur has quit IRC15:17
*** markvoelker has joined #openstack-infra15:17
*** markvoelker has quit IRC15:18
*** jamesmcarthur has joined #openstack-infra15:18
corvusfungi, tristanC: yep,  i think this is just a safety limit for cherrypy.  i believe the chunked upload is working as expected (ie, we're not filling up memory or anything).  it's probably just a default that prevents abuse of "typical" web apps.  and we should raise it very high.15:18
tristanChttp://paste.openstack.org/show/781182/15:18
*** hwoarang has quit IRC15:20
fungiyeah, their tests in cheroot have examples of setting wsgi_server.max_request_body_size15:20
fungiwhich propagates to there15:20
corvuswe can also tell the client what size we would like15:20
corvusthe cherrypy default is 100M15:20
corvusmaybe we should stick with that and use the registry protocol to split further?15:21
*** hwoarang has joined #openstack-infra15:21
fungithat also seems reasonable. 100mb seems like a fairly large chunk size already15:21
openstackgerritTristan Cacqueray proposed zuul/zuul-registry master: Increate max request body size to accept large entity  https://review.opendev.org/68689715:23
corvus(also, this is weird because they both use the same terms -- but the registry protocol has clients uploading "chunks" which are layers split up into multiple http requests -- each of those http requests are actually https chunked transfer encoded, so they are streamed efficiently; it's really the first part of this which is at issue)15:23
tristanChttps://review.opendev.org/686897 fixed my local reproducer15:23
corvusfungi, tristanC: ^ i'm in favor of that for now, and next week working on the docker protocol side and maybe reducing it15:23
fungiyeah, i was using the term "chunk" in both senses15:24
corvushttps://docs.docker.com/registry/spec/api/  "Chunked upload" and "416 Requested Range Not Satisfiable" are the bits i'm looking at for that15:24
openstackgerritJames E. Blair proposed zuul/zuul-registry master: Add debug env variable  https://review.opendev.org/68689515:28
tristanCcorvus: another solution would be to use cherrypy.tools and manage the socket.read manually15:28
tristanCi had to do that to implement the kubectl exec protocol, and cherrypy let you take over easily the low level operation of the request socket15:29
corvustristanC: i don't think we need to do that -- it looks like cherrypy is doing almost exactly what we would do anyway -- in fact, we should drop our 4096 argument to read, then we'll get whole chunks back which we can write to the backend15:30
corvusno, hrm, i'm wrong about that... let's also look at that next week15:32
corvustristanC: maybe you're right and it would be more efficient for us to just grab the socket15:33
corvustristanC: can you +3 https://review.opendev.org/686895 ?15:33
*** Wryhder has joined #openstack-infra15:34
corvusi think everything is approved, so in like an hour or 2, we should recheck that15:34
*** Lucas_Gray has quit IRC15:35
corvusmeanwhile, i'm going to do go whatever it was i was about to do when i sat down here earlier :)15:35
*** Wryhder is now known as Lucas_Gray15:35
fungiyep, i need to stop procrastinating on overdue lawn care this morning, but will be around15:36
tristanCcorvus: cherrypy doesn't seem to read the whole content in our back, but if it does then we could take over the socket15:36
tristanChave a good week end folks :-)15:36
corvustristanC: you too, thanks!15:36
openstackgerritMerged zuul/zuul-registry master: Increate max request body size to accept large entity  https://review.opendev.org/68689715:37
*** tosky has quit IRC15:49
openstackgerritMerged zuul/zuul-registry master: Add debug env variable  https://review.opendev.org/68689515:49
fungi#status log trove instance Etherpad-Dev-MySQL was shut down temporarily by the provider for a proactive cold migration due to an impending host failure, trouble ticket #191005-ord-000033116:00
openstackstatusfungi: finished logging16:00
*** jamesmcarthur has quit IRC16:02
*** jamesmcarthur has joined #openstack-infra16:04
*** icarusfactor has joined #openstack-infra16:06
*** jamesmcarthur has quit IRC16:08
openstackgerritMerged opendev/system-config master: Set debug logging in intermediate registry  https://review.opendev.org/68689616:17
*** odicha has quit IRC16:19
*** ricolin_ has quit IRC16:20
*** jamesmcarthur has joined #openstack-infra16:43
*** dave-mccowan has quit IRC16:43
AJaegerzuul-build-image is working in https://review.opendev.org/686853 now16:43
fungiyay!16:44
AJaegerbut why is it not entering gate pipeline again after a recheck? Bug in Zuul?16:45
AJaegeror did we configure the Zuul tenant wrongly (I think I saw this already in zuul-jobs but didn't dig deeper)16:45
AJaegerfungi, shouldn't it enter gate without your extra +A?16:46
funginot sure, but it got enqueued as soon as i added a new workflow +116:46
AJaegersomething is strange with zuul tenant, with rechecks - in openstack tenant, it would enqueue automatically, seems it's not working in zuul tenant. corvus, any idea?16:47
*** jamesmcarthur has quit IRC16:48
*** jamesmcarthur has joined #openstack-infra17:01
*** jamesmcarthur has quit IRC17:06
*** pcaruana has quit IRC17:18
*** markvoelker has joined #openstack-infra17:19
*** markvoelker has quit IRC17:24
openstackgerritMerged zuul/zuul master: Get ansible version at start of job  https://review.opendev.org/68685317:30
*** e0ne has joined #openstack-infra17:33
*** e0ne has quit IRC17:36
*** slaweq has quit IRC17:44
*** diablo_rojo has joined #openstack-infra18:20
*** goldyfruit_ has joined #openstack-infra18:25
*** yamamoto has joined #openstack-infra19:07
*** yamamoto has quit IRC19:12
*** Lucas_Gray has quit IRC19:33
*** diablo_rojo has quit IRC20:18
*** mattmceuen has joined #openstack-infra20:18
mattmceueno/  whenever anyone gets a chance, could you please add me as a seed member of the airshipapis-core group?   https://review.opendev.org/#/admin/groups/2066,members20:20
mattmceuenthanks!20:20
fungimattmceuen: done!20:21
mattmceuenty fungi!20:21
fungiany time20:21
*** e0ne has joined #openstack-infra20:52
*** eernst_ has joined #openstack-infra21:00
*** xenos76 has quit IRC21:01
*** e0ne has quit IRC21:02
*** markvoelker has joined #openstack-infra21:22
*** markvoelker has quit IRC21:26
*** icarusfactor has quit IRC22:02
*** yamamoto has joined #openstack-infra22:39
*** yamamoto has quit IRC22:48
*** yamamoto has joined #openstack-infra22:51
*** yamamoto has quit IRC22:57
*** jamesmcarthur has joined #openstack-infra22:59
*** jamesmcarthur has quit IRC23:03
*** jamesmcarthur has joined #openstack-infra23:04
*** jamesmcarthur has quit IRC23:08
*** slittle1 has quit IRC23:16
*** slittle1 has joined #openstack-infra23:20
*** jamesmcarthur has joined #openstack-infra23:22
*** yamamoto has joined #openstack-infra23:23
*** yamamoto has quit IRC23:23
*** yamamoto has joined #openstack-infra23:24
*** yamamoto has quit IRC23:28
*** slittle1 has quit IRC23:30
*** slittle1 has joined #openstack-infra23:37
*** slittle1 has quit IRC23:40
*** jamesmcarthur has quit IRC23:49
*** jamesmcarthur has joined #openstack-infra23:51

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!