| *** xinliang has quit IRC | 00:22 | |
| clarkb | pabelanger: reading that code a bit more closely. could part of the problem be not using super? | 00:35 |
|---|---|---|
| clarkb | it is weird to me that we call the methods on the class directly and pass in self rather than using super | 00:36 |
| clarkb | but this could potentially bypass some mro that is required to have that work properly? | 00:36 |
| *** xinliang has joined #zuul | 00:39 | |
| *** xinliang has quit IRC | 00:39 | |
| *** xinliang has joined #zuul | 00:39 | |
| clarkb | pabelanger: there is apparently also IPV6_V6ONLY socket option which if set would require explicit v4 binding too | 00:42 |
| clarkb | but default is to not hav that enabled so I doubt that is our problem | 00:42 |
| clarkb | pabelanger: probably a good start is to update things to use super proper so that mro is respected | 00:45 |
| clarkb | and if it still doesn't work you may need to dig through pdb to see why it is making those choices | 00:45 |
| *** JasonCL has quit IRC | 00:59 | |
| *** JasonCL has joined #zuul | 01:00 | |
| *** JasonCL has quit IRC | 01:00 | |
| *** JasonCL has joined #zuul | 01:01 | |
| *** harlowja has quit IRC | 01:10 | |
| clarkb | https://arstechnica.com/gadgets/2018/03/building-windows-4-million-commits-10-million-work-items/ | 01:26 |
| clarkb | a lot of that feels familiar | 01:26 |
| *** JasonCL has quit IRC | 01:33 | |
| *** JasonCL has joined #zuul | 01:33 | |
| *** JasonCL has quit IRC | 01:37 | |
| *** rlandy|bbl is now known as rlandy | 02:09 | |
| *** myoung|bbl is now known as myoung | 02:52 | |
| *** yolanda has quit IRC | 03:05 | |
| *** JasonCL has joined #zuul | 03:06 | |
| *** JasonCL has quit IRC | 03:11 | |
| *** myoung is now known as myoung|afk | 03:22 | |
| *** rlandy has quit IRC | 03:22 | |
| *** JasonCL has joined #zuul | 03:28 | |
| *** JasonCL has quit IRC | 03:32 | |
| *** harlowja has joined #zuul | 03:48 | |
| *** dtruong2 has joined #zuul | 04:04 | |
| *** yolanda has joined #zuul | 04:05 | |
| *** dtruong2 has quit IRC | 04:08 | |
| *** dtruong2 has joined #zuul | 04:15 | |
| *** dtruong2 has quit IRC | 04:17 | |
| *** pwhalen has quit IRC | 04:42 | |
| *** pwhalen has joined #zuul | 04:42 | |
| *** pwhalen has joined #zuul | 04:42 | |
| *** harlowja has quit IRC | 05:59 | |
| tobiash | abelur: yes, it's completely possible to run the env within docker, but you have to keep in mind a few things | 06:14 |
| tobiash | there are no official docker images yet, so you'll have to build your own | 06:14 |
| tobiash | and for running the zuul-executor bubblewrap you either need unprivileged user namespaces available in your container or need a privileged container | 06:16 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Refactor NodeLauncher to be generic https://review.openstack.org/535555 | 07:09 |
| abelur | tobiash: do you have any pointers? | 07:28 |
| tobiash | abelur: you could use https://review.openstack.org/#/c/465912/ as a starting point | 07:29 |
| abelur | tobiash: thanks ... will give it a shot | 07:31 |
| tobiash | abelur: but that still has some flaws probably so don't expect a first time success ;) | 07:32 |
| tobiash | abelur: unfortunately atm I cannot supply my current docker files because they are taylored to my specific environment | 07:33 |
| tobiash | abelur: but that may change in the future as we already discussed the possibility of supplying official zuul images some time in the future | 07:34 |
| abelur | tobiash: sorry, what does tailored to your env mean? getting it to work with your cloud provider | 07:36 |
| *** hashar has joined #zuul | 07:37 | |
| tobiash | no, taylored to my specific rollout and build process in my private openshift | 07:38 |
| tobiash | with intermediate layers | 07:38 |
| tobiash | like config generation tools | 07:39 |
| abelur | ah ok | 07:39 |
| *** jpena|off is now known as jpena | 08:16 | |
| *** yolanda has quit IRC | 08:35 | |
| *** yolanda_ has joined #zuul | 08:35 | |
| *** JasonCL has joined #zuul | 08:40 | |
| *** Wei_Liu has joined #zuul | 08:45 | |
| *** JasonCL has quit IRC | 08:45 | |
| Wei_Liu | Hello, I want to show something as same as zuul log url in the gerrit, How can I do it? | 08:47 |
| *** sshnaidm|afk is now known as sshnaidm | 08:55 | |
| *** JasonCL has joined #zuul | 09:19 | |
| *** JasonCL has quit IRC | 09:25 | |
| *** sshnaidm has quit IRC | 09:42 | |
| openstackgerrit | Merged openstack-infra/zuul master: Fix error in test_jobs_executed https://review.openstack.org/552142 | 09:43 |
| *** dmellado has joined #zuul | 09:47 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Refactor run_handler to be generic https://review.openstack.org/535554 | 09:49 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Refactor NodeLauncher to be generic https://review.openstack.org/535555 | 09:49 |
| Wei_Liu | Hello, I want to show something as same as zuul log url in the gerrit, How can I do it? | 09:50 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Implement an OpenContainer driver https://review.openstack.org/535556 | 09:53 |
| *** JasonCL has joined #zuul | 10:13 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Implement a Kubernetes driver https://review.openstack.org/535557 | 10:19 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/nodepool master: Implement an Amazon EC2 driver https://review.openstack.org/535558 | 10:22 |
| Wei_Liu | Hello, I want to show something as same as zuul log url in the gerrit, How can I do it? | 10:46 |
| *** sshnaidm has joined #zuul | 10:51 | |
| *** sshnaidm has quit IRC | 11:06 | |
| *** sshnaidm has joined #zuul | 11:21 | |
| *** dkranz has joined #zuul | 11:22 | |
| *** jpena is now known as jpena|lunch | 12:10 | |
| *** myoung|afk is now known as myoung|rover | 12:28 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Make sure we fail because of path checks https://review.openstack.org/552502 | 12:32 |
| tobiash | corvus, clarkb: that improves the tox-remote test ^ | 12:32 |
| *** rlandy has joined #zuul | 12:35 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Make sure we fail because of path checks https://review.openstack.org/552502 | 12:38 |
| *** odyssey4me has quit IRC | 12:41 | |
| *** odyssey4me has joined #zuul | 12:41 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Make sure we fail because of path checks https://review.openstack.org/552502 | 12:47 |
| *** dtruong2 has joined #zuul | 12:58 | |
| *** jpena|lunch is now known as jpena | 13:02 | |
| *** exribe has joined #zuul | 13:26 | |
| *** exribe has quit IRC | 13:28 | |
| *** pwhalen has quit IRC | 13:29 | |
| *** pwhalen has joined #zuul | 13:34 | |
| *** pwhalen has joined #zuul | 13:34 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Make sure we fail because of path checks https://review.openstack.org/552502 | 13:43 |
| *** myoung|rover is now known as myoung|rover|brb | 13:45 | |
| *** myoung|rover|brb is now known as myoung|rover | 13:49 | |
| *** dtruong2 has quit IRC | 13:52 | |
| *** hashar is now known as hasharAway | 14:10 | |
| corvus | Wei_Liu: do you mean you want to make a comment in gerrit that looks like the comments that zuul leaves, but without using zuul? | 14:11 |
| corvus | kklimonda: regarding your question yesterday about two zuul's gating -- an important reason for not doing that is that when zuul gates a project with a dependent pipeline, it tests a change exactly in the way it's going to land, meaning that any changes to the involved projects since the earlier 'check' test are reflected in the 'gate' test (this may even include other in-flight changes). a second zuul | 14:16 |
| corvus | can't participate in that, so it's possible to merge a change which passes the gate jobs in zuul #1, but would not have passed newer check jobs in zuul #2. once that change lands, no further check jobs from zuul #2 would pass, and if zuul #2 is a hard requirement for zuul #1, the system is wedged. | 14:16 |
| corvus | kklimonda: this is why, in openstack, all of the third-party ci zuuls report advisory votes only. the primary zuul ignores them for gating purposes -- they only provide additional info for humans. | 14:17 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: WIP: Handle ZK session loss during node launch https://review.openstack.org/552538 | 14:58 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/nodepool master: WIP: (alt) Handle ZK session loss during node launch https://review.openstack.org/552551 | 15:18 |
| pabelanger | clarkb: super() works, but same exception: socket.gaierror: [Errno -9] Address family for hostname not supported | 15:22 |
| clarkb | ok so likely not a MRO problem then | 15:22 |
| pabelanger | going to see if gdb shows any more info | 15:22 |
| *** harlowja has joined #zuul | 15:46 | |
| clarkb | pabelanger: oh I wonder, it using the hostname to lookup the address in /etc/hosts ? | 15:51 |
| pabelanger | clarkb: not the best at GDB, so can't get much more info. But more reading indicates, python seems to support ipv6, but the interface doesn't. But still is confusing why passing an ipv4 address exposes the failure | 15:52 |
| clarkb | pabelanger: if there isn't a ::1 or similar address in there maybe it is breaking? | 15:52 |
| pabelanger | http://paste.openstack.org/show/700028/ | 15:52 |
| clarkb | pabelanger: in that case if it is using the hostname and not localhost then I could see it failing because it is getting ipv4 and if the ipv4 addr isn't in the ffff:compat format | 15:53 |
| pabelanger | fingergw.listen_address is currently 127.0.0.1, so I don't expect any dns to be used | 15:54 |
| pabelanger | even using 149.202.187.48 still fails | 15:55 |
| clarkb | pabelanger: using python interpreter you can reproduce by creating a socket object? | 15:56 |
| pabelanger | 1min | 15:57 |
| clarkb | might help to remove the tcpserver from the equation and minimize the reproduction | 15:57 |
| pabelanger | clarkb: yah, using example at https://docs.python.org/3.6/library/socket.html#example same exception | 16:00 |
| pabelanger | HOST = '127.0.0.1' | 16:00 |
| pabelanger | HOST = '::1' | 16:01 |
| pabelanger | works | 16:01 |
| clarkb | 'localhost' I assume should work as well since it will be found to be ::1 in /etc/hosts | 16:01 |
| clarkb | pabelanger: my understanding is that to use ipv4 addrs with AF_INET6 you have to use the compat version of an ipv4 addr in ipv6 bitsizing | 16:02 |
| clarkb | so this isn't entirely suprising. I am not sure why it would work elsewhere /me tests locally | 16:02 |
| pabelanger | clarkb: yah, 'localhost' works because resolves to ::1 | 16:03 |
| clarkb | I can reproduce it locally | 16:03 |
| clarkb | maybe the config is localhost sometimes and otehr times not? | 16:04 |
| *** harlowja has quit IRC | 16:06 | |
| pabelanger | clarkb: I assume something like https://review.openstack.org/551015/ works locally for you? | 16:11 |
| clarkb | pabelanger: it should because it should find AF_INET for 127.0.0.1 instead of AF_INET6 | 16:11 |
| clarkb | I'm currently hardcoding AF_INET6 to reproduce the old behavior | 16:11 |
| SpamapS | hrm | 16:33 |
| SpamapS | http://paste.openstack.org/show/700077/ .... can't figure out where the evidence is that this change did not merge | 16:33 |
| corvus | SpamapS: it ran no jobs; that's considered a failure, so it reported and did not merge the pr. | 16:41 |
| corvus | SpamapS: 2018-03-12 21:43:46,769 DEBUG zuul.Pipeline.GoDaddy.gate: No jobs for change <Change 0x7efcb54742e8 890,1309950d972a9de4e90b1b3f35d46f18c54f7375> | 16:41 |
| corvus | (that suggests that should not be a debug line :) | 16:42 |
| *** myoung|rover is now known as myoung|rover|brb | 16:47 | |
| clarkb | pabelanger: fwiw I think your change makes sense because we can't assume everyone will have working ipv6. But we should possibly also consider switching to using localhost and letting name resolution decide what that means | 16:49 |
| tobiash | SpamapS: a do you work with file filters? | 16:49 |
| tobiash | I've seen this often if every job is filtered out of the gate | 16:50 |
| *** myoung|rover|brb is now known as myoung|rover | 16:50 | |
| tobiash | In thaz case just run at least one job without filters or add noop | 16:50 |
| pabelanger | clarkb: switch to localhost for fingergw.listen_address? | 16:51 |
| clarkb | pabelanger: ya | 16:51 |
| clarkb | assuming we use 127.0.0.1 today | 16:51 |
| pabelanger | clarkb: so, I initially did that in testing, but does mean we add a dependency for DNS to setup fingergw, which might be okay. But, IIRC, no other listen_address in zuul.conf would have that requirement. | 16:52 |
| clarkb | pabelanger: there is sort of an implied name service dependency for any internetworking service | 16:53 |
| clarkb | pabelanger: even if it is just a hosts file | 16:53 |
| clarkb | pabelanger: I don't think we would make it a requirement either. I think we also merge your change | 16:53 |
| pabelanger | yah, I can see that | 16:53 |
| SpamapS | tobiash: yes | 16:53 |
| SpamapS | but I have a noop job for that case | 16:53 |
| pabelanger | clarkb: cool, lets see what tobiash, Shrews or corvus says about https://review.openstack.org/551015/ | 16:53 |
| tobiash | pabelanger: what's the expected structure of args? | 16:57 |
| pabelanger | tobiash: tuple with (host, port) | 16:59 |
| tobiash | So the first arg is a tuple? | 17:00 |
| pabelanger | tobiash: also, open to suggestions to make that cleaner | 17:00 |
| pabelanger | yah, arg[0] | 17:00 |
| pabelanger | args[0] | 17:00 |
| tobiash | pabelanger: mind adding a short comment? | 17:01 |
| pabelanger | sure | 17:01 |
| pabelanger | tobiash: just explaining what args[0] is? | 17:01 |
| tobiash | Yah | 17:01 |
| tobiash | Lgtm otherwise | 17:02 |
| *** myoung|rover is now known as myoung|food | 17:04 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 17:06 |
| pabelanger | done | 17:06 |
| tobiash | Nit: typo: truple | 17:10 |
| pabelanger | doh | 17:14 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 17:14 |
| pabelanger | tobiash: fixed^ | 17:14 |
| kklimonda | corvus: both "no jobs for change" and "dependency cycle detected" should probably be reported back to the source - right now users have no feedback on what's happening and they can only guess (or rather ask admins) | 17:14 |
| clarkb | kklimonda: there is a bug open at least for the dependency cycle reporting | 17:15 |
| *** sshnaidm has quit IRC | 17:16 | |
| corvus | kklimonda: yeah, we're getting closer to being able to report dependency cycles with the new depends-on handling. still a bit more work to do. | 17:19 |
| tobiash | Yah, my users also were not only one at my desk due to a dep cycle ;) | 17:19 |
| corvus | no-jobs is trickier | 17:19 |
| corvus | (as in, it needs more thought about exactly when it should or should not report something) | 17:20 |
| corvus | for example, in openstack-infra, we definitely do not want to report every time we don't run jobs on ansible/ansible. | 17:20 |
| tobiash | That would generate many new friends... | 17:21 |
| mrhillsman | is it possible to submit a local directory as a git source? | 17:23 |
| clarkb | mrhillsman: no | 17:23 |
| mrhillsman | thx | 17:23 |
| clarkb | mrhillsman: you can put empty files in it though which is common practice then add the empty file and push that | 17:23 |
| clarkb | or maybe I misread what you were asking | 17:23 |
| clarkb | realizing I'm now in #zuul this may mean use local git repo in zuul configs? I Think you can do that | 17:24 |
| corvus | clarkb, mrhillsman: yeah, if clarkb's revised interpretation is correct, the answer is yes. you can use a file:/// url with the git driver | 17:24 |
| mrhillsman | i think that is what i am looking for | 17:24 |
| tobiash | A local git repo would constrain you to a single host zuul deployment though | 17:25 |
| corvus | (that was actually the first thing the git driver was created for -- long ago i needed a way to actually start zuulv3 without a gerrit :) | 17:25 |
| tobiash | Or be hard to sync | 17:25 |
| corvus | yes -- for production use, it'd be better to use a git repo on the network at least | 17:25 |
| corvus | but for local testing, it can be handy | 17:26 |
| tobiash | Definitely | 17:26 |
| mrhillsman | i'd like to be able to like pull down a repo and without a PR or direct commit trigger ci run with what i have done locally | 17:27 |
| mrhillsman | with what i have locally being used in the run | 17:28 |
| corvus | mrhillsman: the trigger part of that is tricky with the git driver. it doesn't really support 'changes' at all, it can only deal with branch tips. it's better for semi-static configuration (eg, something like the zuul-jobs repo), but, at the moment at least, can't really handle evaluating proposed changes. | 17:29 |
| mrhillsman | ok thx | 17:30 |
| *** jpena is now known as jpena|off | 17:38 | |
| pabelanger | clarkb: mind leaving a review on https://review.openstack.org/551015/ when you have time | 17:45 |
| corvus | FYI: https://review.openstack.org/552637 Remove Zuul from OpenStack governance | 17:47 |
| clarkb | pabelanger: we might also want to update zuul/ansible/library/zuul_console.py at some point too but I guess that is likely going away in the future? | 17:47 |
| clarkb | pabelanger: it hardcodes :: so I think it will only break if ipv6 isn't enabled on a host | 17:47 |
| clarkb | corvus: do you want to review https://review.openstack.org/#/c/551015/7 before it gets approved? | 17:48 |
| corvus | yeah, mordred's streaming changes should remove that | 17:48 |
| pabelanger | clarkb: oh, neat. Didn't know that, or testing hasn't exposed that as a failure | 17:48 |
| pabelanger | kk | 17:48 |
| Shrews | corvus: just looking at that review actually... unix socket forwarding? wha? | 17:49 |
| corvus | Shrews: i know! it's a thing! it's crazy! it's new. | 17:49 |
| * Shrews grepping the googles now | 17:49 | |
| corvus | clarkb, pabelanger: help me out here -- that's the finger daemon? | 17:50 |
| corvus | Shrews: you may be interested in https://review.openstack.org/551015 | 17:51 |
| pabelanger | corvus: clarkb: yah, the exception is from zuul-fingergw | 17:51 |
| pabelanger | http://logs.openstack.org/34/550834/8/check/windmill-ubuntu-bionic/761298f/logs/zf01/var/log/journal/zuul-fingergw.service.log for traceback | 17:51 |
| corvus | clarkb, pabelanger: how did we discover that problem? we don't run that on an ovh host in infra... | 17:51 |
| Shrews | yep, saw that. i'm still confused on why it's needed (what's unique about ovh?), but if it fixes something w/o breaking other things, i'm fine with it | 17:52 |
| corvus | oooh, i see, this came up in deployment functional tests... | 17:52 |
| pabelanger | corvus: yah | 17:52 |
| corvus | how come it hasn't shown up in unit tests? | 17:52 |
| Shrews | i think our unit tests use ::1 for connecting | 17:52 |
| Shrews | iirc | 17:53 |
| clarkb | my understanding is that if you want to hardcode AF_INET6 and use ipv4 addresses you have to provide them in the ipv6 compat format | 17:53 |
| pabelanger | we only bind to ::1 for unit tests I believe, but agree we should write a test to expose it | 17:53 |
| clarkb | so it breaks when using 127.0.0.1 | 17:53 |
| clarkb | and I don't think this will be restricted to ovh, it breaks on my local machine too | 17:53 |
| pabelanger | I've been trying to write up something for that | 17:53 |
| corvus | so put another way, the issue is if you specify "127.0.0.1" instead of "::1" in the config file? | 17:54 |
| clarkb | corvus: yes | 17:54 |
| Shrews | clarkb: orly? do you have test code i can steal? | 17:54 |
| *** sshnaidm has joined #zuul | 17:56 | |
| clarkb | Shrews: http://paste.openstack.org/show/700183/ | 17:56 |
| Shrews | hrm, our func test uses 'localhost' | 17:56 |
| Shrews | clarkb: thx | 17:56 |
| clarkb | localhost will work as long as /etc/hosts has an ::1 entry | 17:56 |
| clarkb | " The address notation for IPv6 is a group of 8 4-digit hexadecimal numbers, separated with a ':'. "::" stands for a string of 0 bits. Special addresses are ::1 for loopback and ::FFFF:<IPv4 address> for IPv4-mapped-on-IPv6." | 17:58 |
| clarkb | the ipv4 mapped on ipv6 addresses are what you need to use to have working ipv4 on an AF_INET6 socket | 17:58 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: DNM - expose [Errno -9] Address family for hostname not supported https://review.openstack.org/552645 | 17:58 |
| corvus | pabelanger: i like the test. strictly speaking, you're testing listing on all addresess in v6, and listening on localhost in v4. that's probably fine, just wanted to point out the discrepancy. | 18:00 |
| pabelanger | ^exposed failure on local laptop and 551015 does fix it. If people are okay with test, I can update 551015 to include it | 18:00 |
| corvus | (::1 is localhost in v6; 0.0.0.0 is all addresses in v4) | 18:00 |
| *** myoung|food is now known as myoung | 18:00 | |
| *** myoung is now known as myoung|rover | 18:00 | |
| pabelanger | corvus: yah, I can also update that to ::1 if we'd like | 18:01 |
| corvus | i think i understand the change now, and it makes sense to me. +2. i'd like to have the test in there. followup patch is okay. | 18:02 |
| pabelanger | okay, rebasing now | 18:02 |
| Shrews | for reference, here is how the python socket lib handles it in create_connection(): https://github.com/python/cpython/blob/b6e43af669f61a37a29d8ff0785455108e6bc29d/Lib/socket.py#L707 | 18:04 |
| Shrews | which seems to be similar to what pabelanger is doing | 18:04 |
| Shrews | pabelanger: what's the first item in the list returned from getaddrinfo()? is it always the same? | 18:06 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add test for fingergw on ipv4 / ipv6 addresses https://review.openstack.org/552645 | 18:07 |
| pabelanger | Shrews: let me check | 18:07 |
| Shrews | https://docs.python.org/3/library/socket.html#socket.getaddrinfo seems to indicate it can return both v4 and v6 info. | 18:08 |
| Shrews | how do we know the first element is the one to use? | 18:08 |
| corvus | if you give it an ip address, the family should be the same on all the returned values | 18:11 |
| Shrews | what if a hostname is given? | 18:11 |
| SpamapS | corvus: thanks for looking. I had, in fact, forgotten the noop job on gate. | 18:11 |
| corvus | (ie, it will give you AF_INET + TCP, AF_INET + UDP, etc) | 18:11 |
| Shrews | the create_connection() code loops through trying each one | 18:11 |
| pabelanger | Shrews: http://paste.openstack.org/show/700211/ is using an IP address, which confirms is the same as corvus mentioned | 18:12 |
| corvus | Shrews: you'll get what's in dns. i don't know what order. | 18:12 |
| pabelanger | let me do hostname | 18:12 |
| SpamapS | and in fact found that we have another check jobs that are reliable enough to run in gate now. | 18:12 |
| SpamapS | i are a english speakers | 18:12 |
| pabelanger | Shrews: http://paste.openstack.org/show/700216/ is hostname | 18:13 |
| corvus | does tcpsocketserver accept a hostname? | 18:13 |
| clarkb | I want to say ipv6 should always go first | 18:14 |
| clarkb | because you are supposed to fall back on ipv4 if ipv6 fails according to some rfc | 18:14 |
| pabelanger | Shrews: however, to your original question, I admit I've used the logic from nodepool: http://git.openstack.org/cgit/openstack-infra/nodepool/tree/nodepool/nodeutils.py#n50 | 18:14 |
| pabelanger | which always uses the first item in the list | 18:15 |
| corvus | clarkb: well, getaddrinfo seems to return v4 first for me. i don't know if that's reliable. | 18:15 |
| Shrews | corvus: it accepts "localhost" so i'm assuming hostnames are supported by tcpsocketserver | 18:16 |
| corvus | Shrews: i guess we could check each one, and use v6 if we see it, otherwise v4. | 18:16 |
| Shrews | corvus: that sounds safer to me | 18:17 |
| corvus | or, you know, max([x[0] for x in getaddrinfo()]) | 18:17 |
| corvus | (don't actually do that ^) | 18:17 |
| corvus | Shrews: yeah, i think that's the safest thing we can do. afaict, socket.bind will accept a hostname (and not just localhost) | 18:18 |
| corvus | so i think we either do that, or we disallow hostnames ourselves. | 18:18 |
| clarkb | I don't think we should disallow hostnames/fqdns | 18:19 |
| clarkb | if its valid in name resolution it should just work | 18:19 |
| corvus | pabelanger, Shrews, clarkb: okay witched to -1 on 551015 with comment | 18:20 |
| * corvus afks to prepare for lunchtime meeting | 18:20 | |
| Shrews | then again, this is the 'listen_address' value... would it ever really be the hostname? | 18:20 |
| Shrews | meh, maybe i guess | 18:21 |
| clarkb | Shrews: it might if you have many interfaces and want to more easily identify a specific one | 18:21 |
| clarkb | rather than providing a specific ip addr for that interface | 18:21 |
| Shrews | yeah | 18:22 |
| pabelanger | corvus: Shrews: okay, so if we get socket.AF_INET6 back from any item in the list, we exit loop and use it, otherwise if none found, fall back to socket.AF_INET? | 18:23 |
| clarkb | pabelanger: that is how I parsed the conversation above | 18:23 |
| pabelanger | cool | 18:24 |
| Shrews | yeah. i suppose at least one of those must be present | 18:24 |
| clarkb | well if no AF_INET and no AF_INET6 you'll have a hard time running tcp :) | 18:25 |
| pabelanger | yah | 18:27 |
| pabelanger | I'm raising an exception on that, but unsure how that would happen :) | 18:28 |
| Shrews | pabelanger: also, might want to limit results to type=socket.SOCK_STREAM | 18:28 |
| pabelanger | ack | 18:28 |
| clarkb | which should give you an empty list if no inet or inet6 | 18:30 |
| clarkb | since I don't think you can sock stream on anything else? I guess there could be implementations that are weird? | 18:31 |
| Shrews | pabelanger: ooh, better yet, just use type=self.socket_type (https://github.com/python/cpython/blob/3.4/Lib/socketserver.py#L417) | 18:34 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 18:34 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add test for fingergw on ipv4 / ipv6 addresses https://review.openstack.org/552645 | 18:34 |
| pabelanger | Shrews: looking | 18:34 |
| pabelanger | Shrews: for line 82 on https://review.openstack.org/#/c/551015/8/zuul/lib/streamer_utils.py ? | 18:36 |
| Shrews | pabelanger: yes, in the getaddrinfo() call | 18:37 |
| pabelanger | kk | 18:37 |
| Shrews | in case someone does something crazy like decide to make it a UDP server or something | 18:37 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 18:38 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add test for fingergw on ipv4 / ipv6 addresses https://review.openstack.org/552645 | 18:39 |
| pabelanger | okay, how does ^ look | 18:39 |
| Shrews | pabelanger: lgtm, but what is "raise error()"? | 18:40 |
| pabelanger | hmm, 1 sec, my raise syntax might now be correct | 18:40 |
| pabelanger | Shrews: yah, that is copypasta | 18:40 |
| pabelanger | fixing | 18:40 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 18:41 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add test for fingergw on ipv4 / ipv6 addresses https://review.openstack.org/552645 | 18:41 |
| *** ssbarnea has joined #zuul | 20:03 | |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 20:27 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add test for fingergw on ipv4 / ipv6 addresses https://review.openstack.org/552645 | 20:27 |
| pabelanger | Shrews: corvus: tobiash: pep8 fixes :( | 20:28 |
| pabelanger | ^ | 20:28 |
| *** harlowja has joined #zuul | 20:37 | |
| dmsimard | SpamapS: you should join #ara we're talking about your 1000+ host fact run :) | 20:55 |
| jlk | FYI, github3.py just did a 1.0 release! | 20:59 |
| jlk | (yes I know we have a bug w/ it, but upstream is happy to do small bugfix releases if it is indeed a bug in github3 code and not our side) | 21:00 |
| pabelanger | nice | 21:01 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Add zuul_json tests https://review.openstack.org/552688 | 21:03 |
| jlk | and... I was just granted pypi publishing rights to github3.py as well, so I can help with those releases. | 21:04 |
| corvus | jlk: yay! so much good news! | 21:04 |
| SpamapS | jlk: oooo next step, connect it to zuul and make it tag->upload :) | 21:11 |
| jlk | heh | 21:12 |
| pabelanger | zuul-tox-remote seems to only run in check, is that correct? | 21:42 |
| clarkb | pabelanger: iirc yes currently | 21:42 |
| clarkb | we should probably add it to the gate though | 21:42 |
| pabelanger | yah, guess that is what I'm asking, should it also be in gate | 21:43 |
| openstackgerrit | Paul Belanger proposed openstack-infra/zuul master: Add zuul-tox-remote to gate pipeline https://review.openstack.org/552692 | 21:45 |
| *** myoung|rover is now known as myoung|bbl | 21:50 | |
| *** myoung|bbl is now known as myoung|snow|bbl | 21:50 | |
| openstackgerrit | Merged openstack-infra/zuul master: Fix [Errno -9] Address family for hostname not supported https://review.openstack.org/551015 | 21:54 |
| pabelanger | yah, thanks corvus, Shrews, clarkb^ | 21:56 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Add zuul_json tests https://review.openstack.org/552688 | 22:07 |
| ianw | hmm, i typo'd a self depends-on with https://review.openstack.org/#/c/552288 ... zuul doesn't tell me anything. is that expected? | 23:06 |
| clarkb | ianw: its probably treating it as a cycle and we don't report cycles back to gerrit yet | 23:09 |
| clarkb | (there is a bug for that though) | 23:09 |
| ianw | ahh, yeah, it is | 23:14 |
| *** hasharAway has quit IRC | 23:14 | |
| *** rlandy is now known as rlandy|afk | 23:44 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!