*** yolanda_ has joined #zuul | 00:13 | |
*** yolanda has quit IRC | 00:16 | |
ianw | corvus: ok, so yeah it doesn't appear to be any redirect etc magic (which i didn't think it would be, as that all happens in the child, but anyway) --> http://logs.openstack.org/03/567703/2/check/dib-functests-python3/e4f0879/job-output.txt.gz#_2018-05-11_00_15_52_377207 | 00:17 |
---|---|---|
*** harlowja has joined #zuul | 01:06 | |
*** swest has quit IRC | 01:56 | |
*** harlowja has quit IRC | 02:07 | |
*** swest has joined #zuul | 02:10 | |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 02:34 |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 02:43 |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 02:55 |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 03:09 |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 03:15 |
openstackgerrit | Ian Wienand proposed openstack-infra/zuul master: [DNM] Testing for with_items log output https://review.openstack.org/567722 | 03:32 |
*** pcaruana has joined #zuul | 06:32 | |
*** dmellado has quit IRC | 07:09 | |
*** dmellado has joined #zuul | 07:10 | |
*** gtema has joined #zuul | 07:11 | |
openstackgerrit | Artem Goncharov proposed openstack-infra/nodepool master: fail to build image without known type https://review.openstack.org/566437 | 07:25 |
*** jpena|off has quit IRC | 07:40 | |
*** ssbarnea_ has joined #zuul | 07:41 | |
*** ssbarnea_ has quit IRC | 07:49 | |
*** electrofelix has joined #zuul | 08:00 | |
*** jpena has joined #zuul | 08:13 | |
*** gouthamr has quit IRC | 08:34 | |
*** gouthamr has joined #zuul | 08:36 | |
*** openstack has joined #zuul | 09:29 | |
*** ChanServ sets mode: +o openstack | 09:29 | |
*** xinliang has quit IRC | 09:57 | |
*** ssbarnea_ has joined #zuul | 09:57 | |
*** xinliang has joined #zuul | 10:13 | |
*** xinliang has quit IRC | 10:13 | |
*** xinliang has joined #zuul | 10:13 | |
*** bhavik1 has joined #zuul | 10:40 | |
*** bhavik1 has quit IRC | 10:51 | |
*** xinliang has quit IRC | 11:26 | |
*** xinliang has joined #zuul | 11:29 | |
*** xinliang has quit IRC | 11:29 | |
*** xinliang has joined #zuul | 11:29 | |
*** electrofelix has quit IRC | 11:35 | |
*** jpena is now known as jpena|lunch | 11:43 | |
*** jimi|ansible has joined #zuul | 12:20 | |
*** jimi|ansible has joined #zuul | 12:20 | |
*** rlandy has joined #zuul | 12:35 | |
*** ssbarnea_ has quit IRC | 12:43 | |
*** TheJulia is now known as needsleep | 12:47 | |
*** jpena|lunch is now known as jpena | 12:51 | |
*** ssbarnea_ has joined #zuul | 13:12 | |
*** openstackstatus has joined #zuul | 13:13 | |
*** ChanServ sets mode: +v openstackstatus | 13:13 | |
-openstackstatus- NOTICE: Due to a Zuul outage, patches uploaded to Gerrit between 09:00UTC and 12:50UTC, were not properly added to Zuul. Please recheck any patches during this window and apologies for the inconvenience. | 13:17 | |
pabelanger | Trying to figure out why my email to zuul-discuss@lists.zuul-ci.org hasn't showed up yet, sent it almost 20hrs go | 13:32 |
*** openstackgerrit has joined #zuul | 13:42 | |
openstackgerrit | Artem Goncharov proposed openstack-infra/zuul master: dont wait infinitely for the connection to zuul_console server https://review.openstack.org/567861 | 13:43 |
*** elyezer has quit IRC | 13:55 | |
*** elyezer has joined #zuul | 13:58 | |
*** pwhalen has quit IRC | 13:59 | |
openstackgerrit | Artem Goncharov proposed openstack-infra/zuul master: fill `delta` with '0' for `creates` and `removes` command. https://review.openstack.org/567864 | 14:01 |
*** pwhalen has joined #zuul | 14:01 | |
*** pwhalen has joined #zuul | 14:03 | |
*** pwhalen has quit IRC | 14:09 | |
*** pwhalen has joined #zuul | 14:13 | |
*** pwhalen has joined #zuul | 14:13 | |
gtema | any conceptual reasons for not having security_groups in nodepool.openstack.pool? I have difficulties extending default SG due to sharing tenant of zuul with other teams. | 14:23 |
openstackgerrit | Merged openstack-infra/zuul-website master: Add favicon https://review.openstack.org/567685 | 14:25 |
corvus | gtema: no, i think that would be okay. | 14:25 |
gtema | corvus: ok, will work on that the. Thanks | 14:26 |
corvus | clarkb, Shrews, mordred: ^ in case you can think of any issues gtema may run into | 14:26 |
corvus | pabelanger: did you get anywhere with the email issue? | 14:27 |
corvus | pabelanger: i suspect it's because zuul's mailman is not running | 14:28 |
corvus | pabelanger: let's switch to infra for this | 14:28 |
gundalow | corvus: Thanks for the suggestion of removing uuids. https://softwarefactory-project.io/r/#/c/12174/ Hopefully this is right | 14:38 |
gundalow | oh, typo | 14:40 |
mordred | corvus, gtema: nope - I think having support for security groups is a fine idea - we just never did it ourselves because we don't really use security groups for things | 14:48 |
*** openstackgerrit has quit IRC | 14:49 | |
clarkb | corvus: gtema if it is an older cloud then nodepools turning on and of of instances combined with group to group rules hammers the database. If cloud is newer (like last 2.5years or so) this has been fixed | 14:50 |
clarkb | mostly just a warning to be careful with group to group rules | 14:51 |
gtema | clarkb: thanks. Generally it is a "custom" Hauwei public cloud (OpenTelekomCloud, currently on Newton level, but pending upgrade). Security groups are generally working fine. So far I have not seen problems here, but sure - might be the case. | 14:57 |
gtema | any idea why review 567864 might have failed on py36? There is no clear info in logs, just "ft2.17: tests.unit.test_gerrit_crd.TestGerritCRD.test_crd_check_reconfiguration_StringException"? In py35 everything is fine. | 14:59 |
rcarrillocruz | wait, is https://docs.openstack.org/infra/nodepool/configuration.html the logo now? | 15:00 |
rcarrillocruz | nice! | 15:00 |
fungi | rcarrillocruz: indeed it is | 15:01 |
fungi | as seen on https://zuul-ci.org/ | 15:02 |
fungi | even has a page icon as of a few minutes ago | 15:02 |
rcarrillocruz | ++ | 15:03 |
*** pcaruana has quit IRC | 15:07 | |
corvus | gtema: that usually means the test timed out in some way. it probably is not related to that change and you can recheck. the py36 job is new, and may have slightly different timings than the py35 job, so we may encounter some more races in tests for the next while (until we can track them down and fix them) | 15:09 |
corvus | gundalow: yay! looks good | 15:10 |
gtema | corvus: thanks | 15:11 |
Shrews | Does this seem right to anyone else? http://git.openstack.org/cgit/openstack-infra/nodepool/tree/nodepool/driver/__init__.py#n688 | 15:17 |
Shrews | I don't think comparing __dict__ between objects is doing what we might expect it to be doing | 15:17 |
corvus | Shrews: how so? | 15:18 |
Shrews | simple test shows me that contains method pointers and not the actual attributes | 15:18 |
Shrews | B.__dict__ is {'__module__': '__main__', '__init__': <function B.__init__ at 0x7f5175d18840>, '__eq__': <function B.__eq__ at 0x7f5175d188c8>, '__doc__': None, '__hash__': None} | 15:19 |
Shrews | the attributes I defined in that class don't even appear there | 15:19 |
corvus | Shrews: B looks like a class object, not an instance | 15:20 |
Shrews | oh, ha | 15:20 |
Shrews | yep | 15:20 |
Shrews | well now I'm wondering why we keep redefining __eq__ in subclasses | 15:21 |
corvus | Shrews: like where? | 15:21 |
Shrews | corvus: http://git.openstack.org/cgit/openstack-infra/nodepool/tree/nodepool/driver/openstack/config.py#n50 | 15:22 |
Shrews | driver configs, for one | 15:22 |
corvus | Shrews: that's probably "better" :) | 15:23 |
Shrews | corvus: i think so (if they actually compared types before comparing the attrs). | 15:24 |
corvus | if that's universal, we could probably drop __eq__ on the base class; or maybe make it abc. | 15:24 |
corvus | Shrews: heh, yeah they should do that too :) | 15:24 |
Shrews | i'm going to clean this up | 15:24 |
Shrews | otherwise my multi-label patch gets even messier | 15:25 |
*** rlandy is now known as rlandy|brb | 15:40 | |
*** dkranz has quit IRC | 15:49 | |
*** rlandy|brb is now known as rlandy | 16:01 | |
*** acozine1 has joined #zuul | 16:09 | |
*** dkranz has joined #zuul | 16:20 | |
*** dkranz has quit IRC | 16:23 | |
*** dkranz has joined #zuul | 16:25 | |
*** dkranz has quit IRC | 16:27 | |
*** dkranz has joined #zuul | 16:28 | |
*** gtema has quit IRC | 16:53 | |
*** openstackstatus has quit IRC | 17:00 | |
*** openstack has joined #zuul | 17:04 | |
*** ChanServ sets mode: +o openstack | 17:04 | |
tobiash | just did a second try of ansible 2.5 and jobs look good now | 17:14 |
clarkb | tobiash: was the fix for shell and command in the callback the only major fix you needed? | 17:19 |
tobiash | clarkb: that was the only fix needed so far | 17:19 |
mnaser | does zuul have a lot of binary dependencies overall (aka: would there be a lot of issues deploying it on arm?) | 17:20 |
corvus | mnaser: off the top of my head: the zuul executor wants bubblewrap; and there's some crypto stuff in the scheduler | 17:21 |
clarkb | mnaser: the bindep file should give you a good idea of what is needed. re2 is probably the only one I would worry about. bubblewrap too maybe? | 17:21 |
fungi | i think pyca/cryptography deploys fine on arm though | 17:21 |
mnaser | i guess one way to find out | 17:21 |
fungi | as far as having its library dependencies met | 17:21 |
fungi | mnaser: multiple ways to find out, but one fairly easy one would probably be trying to add an arm64 integration test since we have nodes of that description | 17:22 |
*** jpena is now known as jpena|off | 17:22 | |
mnaser | fungi: oh that's true as well, i was going to try deploying it here | 17:22 |
clarkb | fungi: ya cryptography should be fine as evidenced by our arm64 nodepool builder which uses cryptography for image uploads iirc | 17:23 |
corvus | mnaser: pabelanger: could you throw a windmill job on an arm node? | 17:24 |
fungi | that's sorta what i was wondering too | 17:24 |
mnaser | corvus: i pinged pabelanger just now to see if i can get windmill docs to run on a vm but i'm looking at doing it upstream too | 17:24 |
mnaser | :) | 17:24 |
pabelanger | Sure | 17:25 |
pabelanger | 1 minute | 17:25 |
mnaser | i can fix problems as they appear | 17:25 |
corvus | mnaser: i can make problems as you fix them. that's teamwork! | 17:26 |
mnaser | :D | 17:26 |
clarkb | bubblewrap is built for arm64 on ubuntu bionic at least. as is libre2 | 17:26 |
mnaser | maybe this could be the demo | 17:26 |
mnaser | cause i'm supposed to come up with some interesting one.. not sure how long windmill takes to deploy however | 17:26 |
corvus | however long it takes, it will take longer on arm | 17:27 |
mnaser | :\ so my little 8 minutes i'm given wouldn't be enough | 17:27 |
mnaser | (this is supposed to be some high end arm hardware but yeah) | 17:27 |
fungi | if could be a non-live demo with the thumb-twiddling elided? | 17:27 |
mnaser | could be, but that's no fun | 17:28 |
mnaser | :p | 17:28 |
corvus | http://logs.openstack.org/98/564598/1/gate/windmill-ubuntu-bionic/38a6054/job-output.txt.gz took 15 minutes. i have no idea how that's divided up tho | 17:29 |
mnaser | http://logs.openstack.org/83/567883/3/check/windmill-ubuntu-bionic/67cc8e5/logs/ara-report/ | 17:29 |
mnaser | looks like 1m15s just installing zuul | 17:29 |
mnaser | maybe prepopulating cache could help[ | 17:29 |
clarkb | prebuilt wheels likely would help | 17:30 |
clarkb | then hope nothing has a new release on pypi that day (or use a mirror, or tell it no new downloads) | 17:30 |
mnaser | yeah a local mirror just for this | 17:30 |
mnaser | http://zuul.openstack.org/stream.html?uuid=45449af46beb4234970c98e995a7b9c9&logfile=console.log time to watch, thanks for the patch pabelanger | 17:31 |
pabelanger | the test jobs in windill don't use the wheel mirrors in openstack-infra, mostly to confirm we can build things in the gate | 17:31 |
pabelanger | I'm guess it will fail when we get to installing bwrap, unless we created an arm64 package | 17:33 |
pabelanger | we might want to add ubuntu-bionic to arm64 in that case | 17:33 |
clarkb | there is a bionic arm64 package | 17:33 |
*** smyers has quit IRC | 17:33 | |
fungi | of bwrap | 17:34 |
clarkb | yes | 17:34 |
pabelanger | yah, but we don't have that image on linaro | 17:34 |
clarkb | ah | 17:34 |
pabelanger | just xenial and stretch | 17:34 |
corvus | fetch-zuul-cloner ? | 17:34 |
corvus | is this still a legacy job? | 17:34 |
pabelanger | corvus: still happens in our base jobs | 17:34 |
pabelanger | no | 17:35 |
corvus | i thought we only put that in the legacy base job | 17:35 |
pabelanger | removal has been up for some time, but haven't landed it yet | 17:35 |
corvus | i apparently thought wrong. hrm. | 17:35 |
clarkb | I think monty expected much to break if it went in? | 17:35 |
pabelanger | https://review.openstack.org/513506/ | 17:36 |
mnaser | we should add a feature to have # of viewers in a zuul job console | 17:36 |
mnaser | i think we'd see some good numbers on the arm job right now | 17:36 |
mnaser | :P | 17:36 |
*** smyers has joined #zuul | 17:36 | |
corvus | pabelanger, clarkb: how about after summit we try to regroup on that? | 17:37 |
clarkb | wfm | 17:37 |
pabelanger | sure | 17:37 |
corvus | that's several minutes we waste in every job :| | 17:37 |
fungi | soon post-summit sounds like a great time | 17:38 |
clarkb | corvus: isn't it just copying one file? | 17:38 |
clarkb | oh no it has to make a venv right? | 17:38 |
corvus | yep | 17:39 |
corvus | because we removed it from images | 17:39 |
corvus | (which was a step which was supposed to last for about 3 days :) | 17:39 |
clarkb | if only pyyaml were part of stdlib | 17:40 |
fungi | if pyyaml were part of stdlib, pep 518 would never have ended up using toml | 17:41 |
mnaser | c/_cffi_backend.c:15:17: fatal error: ffi.h: No such file or directory | 17:41 |
mnaser | hmmm | 17:41 |
mnaser | is it possible libffi-dev is not being installed? | 17:41 |
corvus | libffi-dev [platform:dpkg] | 17:41 |
corvus | that's from bindep | 17:41 |
mnaser | https://packages.ubuntu.com/xenial/arm64/libffi-dev/download | 17:42 |
mnaser | it also exists.. | 17:42 |
pabelanger | I think I see the issue | 17:42 |
mnaser | bootstrap tries to install bindep | 17:43 |
corvus | where do we run bindep? | 17:43 |
mnaser | so it can run bindep | 17:43 |
mnaser | but it fails in that stage, i don't see bindep running | 17:43 |
pabelanger | tools/install_bindep.sh is what runs it, but that happens after the virtuelenv is created | 17:44 |
mnaser | pabelanger: makes me curious how that is not an issue on other architectures | 17:47 |
pabelanger | mnaser: not sure, will need to debug that in a bit | 17:48 |
*** openstackgerrit has joined #zuul | 17:55 | |
openstackgerrit | Monty Taylor proposed openstack-infra/nodepool master: Use openstacksdk instead of os-client-config https://review.openstack.org/566158 | 17:55 |
pabelanger | I noticed when setting up a 3pci system, the tox jobs doesn't work out of box. This is because DIB images, don't include tox by default, and when run ensure-tox role, it will pip install tox into ~/.local/bin/tox but a user then needs to set tox_executable: ~/.local/bin/tox manually, as .local/bin isn't in the default path | 18:14 |
pabelanger | I was thinking of proposing tox_executable: ~/.local/bin/tox as the default for the tox job because of that | 18:15 |
clarkb | or maybe just add that to the path? | 18:15 |
pabelanger | I am not sure best place to do that | 18:15 |
clarkb | on the shell command running tox? | 18:16 |
pabelanger | right now I have to shadow zuul-jobs to modify the job to just set tox_executable | 18:16 |
clarkb | though I guess if nothing else is installed there then they are equivalent in function | 18:16 |
pabelanger | seems like a lot of work just for 3pci | 18:17 |
clarkb | I would update zuul-jobs | 18:17 |
pabelanger | not sure, we are saying tox_executable is the path to tox in our docs | 18:18 |
*** acozine1 has quit IRC | 18:19 | |
pabelanger | will need to see if command task will accept PATH | 18:19 |
clarkb | oh the tox role isn't installing tox | 18:19 |
pabelanger | right, we have ensure-tox that does it | 18:20 |
pabelanger | this isn't an issue in openstack-infra, because we include tox at build time | 18:20 |
clarkb | can ensure-tox register tox_executable to be ~/.local/bin/tox if it installs tox? | 18:21 |
clarkb | except that runs in pre.yaml | 18:21 |
pabelanger | right | 18:21 |
pabelanger | and I think would make it more tricky if a user setup tox_executable themself | 18:21 |
pabelanger | I'm thinking we set tox_environment: ~/.local/bin/tox by default and in the case of openstack-infra, we maybe set a site-variable (like we did with bindep) | 18:22 |
clarkb | I would have tox check instead | 18:23 |
clarkb | then you shouldn't need to change any config anywhere | 18:23 |
pabelanger | actually, what is we removed tox from infra-packages in openstack-infra, then just used ensure-tox role, like other zuul users would | 18:25 |
pabelanger | then we'd wouldn't need to check for more then 1 location | 18:25 |
clarkb | that assumes we are the only people preinstalling tox | 18:25 |
pabelanger | right, give that we don't have docs yet in zuul to actually build a DIB, we could make ask on ML | 18:26 |
openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix ConfigValue comparisons https://review.openstack.org/567927 | 18:26 |
Shrews | so that ^ was "fun" | 18:26 |
clarkb | or just check if it is in path use `tox` else use ~/.local/bin/tox ? | 18:26 |
clarkb | pabelanger: one hack that can be used is add a symlink to ~/.local/bin/tox to /usr/local/bin in ensure-tox | 18:28 |
clarkb | (this is actually how I use tox locally) | 18:28 |
fungi | yeah, works fine so long as the user invoking tox has read access to whichever homedir that got installed into | 18:30 |
clarkb | Shrews: forgive my ignorance but wouldn't the simplest fix to just be compare __dict__ and remove any __eq__ overrides?? | 18:31 |
Shrews | clarkb: that doesn't take into account class types | 18:31 |
Shrews | so a ConfigValue could be equal to some random dict | 18:31 |
pabelanger | maybe it is just easier to setup PATH=$PATH:$HOME/.local/bin:$HOME/bin for zuul user in dib element | 18:32 |
Shrews | clarkb: but also, there are some attributes we DONT want to compare | 18:32 |
Shrews | clarkb: see the NOTEs i left about recursion | 18:32 |
clarkb | Shrews: in that case maybe check the isinstance then just super to the base class __eq__ ? | 18:32 |
clarkb | oh if you don't want to compare some things then ya need to override | 18:32 |
Shrews | i missed the static driver changes for that :/ | 18:33 |
clarkb | pabelanger: that doesn't fix the 3pci images does it? if so you could just install tox on those images | 18:33 |
pabelanger | clarkb: right, same issue modifying a default image | 18:33 |
pabelanger | okay, I'll look at tox role and see about doing double check | 18:34 |
clarkb | lrwxrwxrwx 1 clark users 24 Oct 3 2016 /home/clark/bin/tox -> /home/clark/venv/bin/tox is my hack | 18:36 |
clarkb | pabelanger: something like register command: bash -c 'PATH=~/.local/bin:$PATH type -p {{tox_executable}}' if rc != 0 error | 18:49 |
corvus | clarkb, pabelanger: yeah, the goals should be that ensure-tox works whether tox is pre-installed or not. and that the tox role works in either case without special arguments (you should only need to set those if you, say, wanted to run tox from some really weird place). | 18:54 |
corvus | clarkb, pabelanger: so it sounds like something like the check clarkb is proposing in the tox role should maybe do the trick | 18:54 |
clarkb | the neat thing about type is that if you give it a full path that is valid it just spits it back at you | 18:56 |
clarkb | so it should be backward compatible with any full path as well as just executable name | 18:57 |
pabelanger | sure, I'll work on a patch for this afternoon | 19:01 |
*** acozine1 has joined #zuul | 19:07 | |
*** gtema has joined #zuul | 19:07 | |
SpamapS | mmmmm website with logo and tagline | 19:09 |
* SpamapS just needs the t-shirt next. | 19:10 | |
SpamapS | and stiiickkarrss | 19:10 |
openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: Fix ConfigValue comparisons https://review.openstack.org/567927 | 19:17 |
*** gtema has quit IRC | 20:09 | |
*** acozine1 has quit IRC | 21:26 | |
*** rlandy has quit IRC | 22:03 | |
*** ssbarnea_ has quit IRC | 22:16 | |
gundalow | zuul-jobs/roles/prepare-workspace/tasks/main.yaml's Synchronize src repos to workspace directory is failing for me, thought the task has no_log: true. Can I sub class that and remove the no_log? https://ansible.softwarefactory-project.io/logs/13/13/793d6e03a7af2d23cb6d333409c4079c63514471/check/ansible-test-network-integration-vyos-devel/0cedf45/ara/ hitting issues with a new node type I'm bring up | 22:39 |
corvus | gundalow: you'll need a new base job that uses a new role instead of the standard prepare-workspace role | 22:50 |
corvus | gundalow: where's your project config repo again? (the repo that had that nodepool change earlier) | 22:52 |
corvus | gundalow: nm, found it: github.com/ansible-network/zuul-config | 22:52 |
gundalow | corvus: https://github.com/ansible-network/ansible-zuul-jobs/pull/13/files#diff-dbd9cd0ce7e9a5628770143f1488ff59R38 defines the job that uses the ansible-network-vyos-1.1.8 | 22:52 |
gundalow | node | 22:52 |
*** smyers has quit IRC | 22:53 | |
gundalow | ansible-network-vyos-1.1.8 = Fedora 27 with zuul user & SF's public key | 22:53 |
corvus | gundalow: want me to write a change real quick that should get you the logs? | 22:53 |
gundalow | corvus: That would be great :) | 22:53 |
corvus | k, just a min | 22:54 |
gundalow | At a guess I think it may be auth problems, or maybe dest directory/permissions are wrong | 22:54 |
corvus | yeah sounds likely | 22:54 |
*** smyers has joined #zuul | 22:55 | |
corvus | gundalow: okay here we go: https://github.com/ansible-network/zuul-config/pull/18 | 23:08 |
corvus | gundalow: now since zuul-config is a config repository, zuul's usual behavior of dynamically altering the config doesn't apply -- we have to actually land that change before we can use it | 23:09 |
corvus | gundalow: but once you do, you can update https://github.com/ansible-network/ansible-zuul-jobs/pull/13 to add "parent: base-test" to the job definition, and it will use that one instead of the normal base job. | 23:10 |
gundalow | Merged the PR | 23:10 |
gundalow | Added https://github.com/ansible-network/ansible-zuul-jobs/pull/13/commits/ce419f41d8b48bbb429a8d039136eea87926b3ac | 23:15 |
corvus | running at https://ansible.softwarefactory-project.io/zuul/stream.html?uuid=6caf6f4ffab748eaa75ba04001b8706b&logfile=console.log | 23:15 |
gundalow | corvus: missing rsync | 23:16 |
corvus | gundalow: yay! it's an answer at least :) | 23:16 |
gundalow | And something I can fix | 23:16 |
gundalow | though not today | 23:16 |
gundalow | corvus: nhicher Thank you again for all your help! | 23:17 |
gundalow | corvus: Not sure where in the world you are, though it should be weekend for you now | 23:18 |
corvus | gundalow: np, have a good weekend :) | 23:18 |
gundalow | Thanks for this though, means I can solve more of my own problems now :) | 23:18 |
corvus | i feel it it should have been weekend yesterday | 23:18 |
gundalow | :D | 23:18 |
corvus | gundalow: also, let me know if you have any questions about the pr i submitted | 23:19 |
corvus | i mean, after weekending | 23:19 |
openstackgerrit | James E. Blair proposed openstack-infra/zuul master: WIP replace use of asyncio with cherrypy https://review.openstack.org/567959 | 23:25 |
corvus | mordred, tristanC, Shrews, tobiash: ^ i needed a break from working on slides, so i took a couple of hours to poke at that ^. i have the unit tests for the basic gearman status check and the websocket console streaming working. | 23:27 |
corvus | mordred, tristanC, Shrews, tobiash: the advantage of that is that we can drop our dependence on aiohttp (which has caused us a number of problems recently as they are providing a very unstable api), and cherrypy is a more traditional pythonic threaded system, so it should fit in better with how the rest of the system works. | 23:29 |
corvus | we can stop typing "await" all the time. | 23:29 |
corvus | it supports a routes-based dispatcher, so the mapping to what we're doing is pretty straightforward. i think porting the rest of the webapp will only take a couple hours. | 23:30 |
clarkb | does it do websockets? that was the original aiohttp/asyncio reason iirc | 23:30 |
corvus | clarkb: ws4py provides a cherrypy plugin to add websocket support | 23:31 |
clarkb | cool | 23:31 |
corvus | locally, i have tests.unit.test_web.TestWeb.test_web_status and tests.unit.test_streaming.TestStreaming.test_websocket_streaming passing | 23:31 |
corvus | they probably won't pass in zuul because i forgot to add the packages to requirements.txt | 23:31 |
corvus | but there's code in that patch which implements the websockets console streaming | 23:32 |
corvus | (another nice thing is that with one more minor refactor, i think we can achieve Shrews's dream of having a single helper class for all our streamers) | 23:33 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!