| *** ansmith has joined #openstack-oslo | 00:00 | |
| *** phuongnh has joined #openstack-oslo | 00:44 | |
| *** ducnv has quit IRC | 01:18 | |
| *** zaneb has joined #openstack-oslo | 01:31 | |
| *** zaneb has quit IRC | 01:31 | |
| *** zaneb has joined #openstack-oslo | 01:31 | |
| *** phuongnh has quit IRC | 01:57 | |
| *** phuongnh has joined #openstack-oslo | 01:58 | |
| *** ianychoi has quit IRC | 02:10 | |
| *** ianychoi has joined #openstack-oslo | 02:10 | |
| *** toabctl has quit IRC | 02:19 | |
| *** toabctl has joined #openstack-oslo | 02:23 | |
| *** ducnv has joined #openstack-oslo | 02:41 | |
| *** lbragstad has quit IRC | 03:16 | |
| *** cfriesen has quit IRC | 03:27 | |
| *** cburgess has quit IRC | 03:41 | |
| *** cburgess has joined #openstack-oslo | 03:41 | |
| *** dave-mccowan has joined #openstack-oslo | 04:28 | |
| *** dave-mccowan has quit IRC | 04:53 | |
| *** prometheanfire has joined #openstack-oslo | 05:58 | |
| prometheanfire | https://review.openstack.org/615724 fixes a nova thing, more reviews would be nice (oslo.service) | 05:58 |
|---|---|---|
| prometheanfire | :D | 05:58 |
| *** e0ne has joined #openstack-oslo | 06:24 | |
| *** e0ne has quit IRC | 06:44 | |
| *** e0ne has joined #openstack-oslo | 06:48 | |
| *** phuongnh has quit IRC | 06:51 | |
| *** phuongnh has joined #openstack-oslo | 06:52 | |
| *** e0ne has quit IRC | 07:07 | |
| *** pcaruana has joined #openstack-oslo | 07:36 | |
| *** cfriesen has joined #openstack-oslo | 07:39 | |
| *** cfriesen has quit IRC | 07:48 | |
| openstackgerrit | wangxiyuan proposed openstack/oslo.limit master: Add basic enforce func https://review.openstack.org/596520 | 08:50 |
| *** mattgo has joined #openstack-oslo | 09:05 | |
| *** mattgo has quit IRC | 09:05 | |
| ttx | bnemec: it's a bit borderline... The cases where that would affect users are probably close to 0... so I could go either way | 09:15 |
| ttx | (x or y) | 09:15 |
| ttx | But I would definitely not include that one in stable branches :) | 09:16 |
| *** cdent has joined #openstack-oslo | 09:22 | |
| *** lbragstad has joined #openstack-oslo | 10:00 | |
| *** phuongnh has quit IRC | 10:42 | |
| openstackgerrit | wanghui proposed openstack/taskflow master: Remove stestr from requirements.txt https://review.openstack.org/615825 | 10:58 |
| openstackgerrit | wanghui proposed openstack/oslotest master: Remove stestr from requirements.txt https://review.openstack.org/615826 | 11:01 |
| *** lbragstad has quit IRC | 11:02 | |
| *** lbragstad has joined #openstack-oslo | 11:04 | |
| *** e0ne has joined #openstack-oslo | 11:11 | |
| *** raildo has joined #openstack-oslo | 11:54 | |
| *** pbourke has quit IRC | 12:11 | |
| *** pbourke has joined #openstack-oslo | 12:11 | |
| *** e0ne has quit IRC | 12:29 | |
| *** dave-mccowan has joined #openstack-oslo | 12:29 | |
| *** ansmith has quit IRC | 12:39 | |
| *** e0ne has joined #openstack-oslo | 12:43 | |
| *** cdent has quit IRC | 13:31 | |
| *** kgiusti has joined #openstack-oslo | 13:37 | |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.service stable/rocky: Fix stop of loopingcall https://review.openstack.org/615263 | 14:00 |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.service stable/rocky: Use eventlet Event for loopingcall events https://review.openstack.org/614489 | 14:03 |
| *** ansmith has joined #openstack-oslo | 14:07 | |
| *** bobh has joined #openstack-oslo | 14:19 | |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.service stable/rocky: Fix stop of loopingcall https://review.openstack.org/615263 | 14:25 |
| openstackgerrit | Merged openstack/oslo.log master: Add Windows Event Log handler https://review.openstack.org/614558 | 14:29 |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.service stable/rocky: Use eventlet Event for loopingcall events https://review.openstack.org/614489 | 14:30 |
| *** munimeha1 has joined #openstack-oslo | 14:58 | |
| *** cfriesen has joined #openstack-oslo | 15:04 | |
| *** e0ne has quit IRC | 15:06 | |
| *** prometheanfire has left #openstack-oslo | 15:09 | |
| *** cdent has joined #openstack-oslo | 15:36 | |
| openstackgerrit | Stephen Finucane proposed openstack/oslo.upgradecheck master: doc: Remove crud from conf.py file https://review.openstack.org/609327 | 16:26 |
| openstackgerrit | Stephen Finucane proposed openstack/oslo.upgradecheck master: doc: Switch to openstackdocstheme https://review.openstack.org/609328 | 16:26 |
| dhellmann | kgiusti : what is our current recommended driver for notifications? rabbit or kafka or something else? | 17:02 |
| kgiusti | dhellmann: rabbitmq | 17:02 |
| dhellmann | I'm considering writing something that consumes notifications, but want to use asyncio so I don't think I'll be able to use oslo.messaging directly (at least in the early stages; I would consider adding asyncio support at some point). Do you think if I went with rabbitmq, that would make it work with most deployments? | 17:03 |
| kgiusti | dhellmann: yes, definitely rabbitmq for that | 17:04 |
| dhellmann | k | 17:04 |
| dhellmann | kgiusti : thanks! | 17:04 |
| kgiusti | dhellmann: yw | 17:05 |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.config master: Introduce a base yaml parser for all openstack components https://review.openstack.org/615529 | 17:18 |
| openstackgerrit | Hervé Beraud proposed openstack/oslo.config master: Introduce a base yaml parser for all openstack components https://review.openstack.org/615529 | 17:20 |
| *** e0ne has joined #openstack-oslo | 17:32 | |
| openstackgerrit | Merged openstack/oslo.service stable/rocky: Fix stop of loopingcall https://review.openstack.org/615263 | 17:34 |
| *** e0ne has quit IRC | 18:03 | |
| bnemec | kmalloc: zzzeek: Any thoughts on how to fix https://bugs.launchpad.net/oslo.cache/+bug/1801967 ? | 18:03 |
| openstack | Launchpad bug 1801967 in oslo.cache "TypeError: unhashable type: 'list' with redis driver" [Critical,Triaged] | 18:03 |
| kmalloc | bnemec: ugh. | 18:04 |
| kmalloc | i think we're going to need to add a "handle arguments" per backend | 18:04 |
| bnemec | Yeah | 18:05 |
| kmalloc | the core of the issue is the backends are opaque | 18:05 |
| kmalloc | so we can't know unless we know what library is being used behind the scenes | 18:05 |
| kmalloc | this is an issue with dogpile not being consistent in it's backends | 18:05 |
| kmalloc | and just letting the underlying library handle it | 18:05 |
| kmalloc | i'll just re-write the memcache backend to be pymemcache and handle the url arguments | 18:06 |
| kmalloc | lets revert my fix | 18:06 |
| kmalloc | i'll overload the backend bits in oslo.cache for now and change the default behaviors there | 18:06 |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache master: Revert "Config dict arguments.url must be a list" https://review.openstack.org/615960 | 18:06 |
| *** e0ne has joined #openstack-oslo | 18:06 | |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache stable/rocky: Revert "Config dict arguments.url must be a list" https://review.openstack.org/615961 | 18:06 |
| kmalloc | ^ | 18:07 |
| kmalloc | i should have a new driver written today to fix the core of that issue. | 18:07 |
| bnemec | Sounds good, thanks | 18:07 |
| kmalloc | but it's much easier to just revert than chase redis fixes down | 18:08 |
| kmalloc | also... someone should chat with ceilometer and tell them they should never use redis unless they're using pre-new license. | 18:08 |
| * kmalloc gets off soapbox on that last one. | 18:09 | |
| *** hberaud has quit IRC | 18:14 | |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache master: Revert "Config dict arguments.url must be a list" https://review.openstack.org/615960 | 18:14 |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache stable/rocky: Revert "Config dict arguments.url must be a list" https://review.openstack.org/615961 | 18:15 |
| *** hberaud has joined #openstack-oslo | 18:15 | |
| kmalloc | bnemec: ^ with proper bug tag | 18:15 |
| *** e0ne has quit IRC | 18:24 | |
| *** efried has joined #openstack-oslo | 18:40 | |
| efried | melwitt, bnemec, dhellmann: So as not to duplicate effort, I'm working on that fixture (https://review.openstack.org/#/c/615724/) | 18:41 |
| melwitt | efried: woot, wfm | 18:42 |
| *** prometheanfire has joined #openstack-oslo | 18:49 | |
| prometheanfire | does osprofiler have an irc channel (or possibly share it here)? | 18:50 |
| kmalloc | prometheanfire: afaik it's pretty dead. | 18:51 |
| kmalloc | prometheanfire: in all of rocky, osprofiler had 35 reviews in the repo | 18:51 |
| prometheanfire | ya | 18:51 |
| prometheanfire | kinda thinking that too | 18:52 |
| kmalloc | prometheanfire: i'm about to rip it out of keystone | 18:52 |
| prometheanfire | I want them to remove jaeger-client from it | 18:52 |
| prometheanfire | they aren't even using it | 18:52 |
| prometheanfire | nvm, git pull, they are using it | 18:52 |
| prometheanfire | :( | 18:53 |
| *** bobh has quit IRC | 18:55 | |
| *** shardy has quit IRC | 19:00 | |
| kmalloc | bnemec: hm. so... | 19:12 |
| kmalloc | bnemec: i need to handle serialization in the new memcache backend.... python-memcached uses pickle | 19:13 |
| kmalloc | but to keep the most consistent behavior, that is the best case. | 19:13 |
| kmalloc | maybe msgpack? | 19:13 |
| bnemec | I know we added msgpack to something else recently. | 19:14 |
| bnemec | Maybe messaging? | 19:14 |
| kmalloc | oslo_serialization uses it | 19:14 |
| kmalloc | but it is not as clean as pickle | 19:14 |
| kmalloc | pickle (sadly) just works | 19:14 |
| kmalloc | msgpack requires some special magic to load some classes back from serialization | 19:15 |
| kmalloc | we found this out in keystone (we use it there) | 19:15 |
| kmalloc | i am inclined to just use pickle for now for consistency and then we do a deprecation cycle on that behavior | 19:15 |
| kmalloc | ? | 19:15 |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Fixture to mock loopingcall wait() https://review.openstack.org/615978 | 19:15 |
| kmalloc | ... i guess i could just re-implement the same backend for now and do pymemcache separately | 19:16 |
| efried | melwitt, bnemec, dhellmann: ---^ | 19:16 |
| kmalloc | it's a 1-line fix. | 19:16 |
| bnemec | kmalloc: I guess I'm not that familiar with the architecture of oslo.cache so I'm not sure I can have an informed opinion on this. | 19:16 |
| * kmalloc wrote most of the core of oslo.cache for keystone :P | 19:16 | |
| kmalloc | so, what happens is there are three tiers of objects: region, proxy, backend | 19:17 |
| kmalloc | region is the main interface, proxy catches data passed to the backend and modifies it | 19:17 |
| kmalloc | and backends are like redis, pylibmc, python-memcached | 19:17 |
| kmalloc | etc | 19:17 |
| kmalloc | and implement to the lowest common/most basic functionality (python-memcached) | 19:18 |
| kmalloc | the issue is python-memcached doesn't understand "host:port, host:port, host:port" and expects [host:port, host:port, host:port] | 19:18 |
| *** prometheanfire has left #openstack-oslo | 19:18 | |
| kmalloc | everything else is fine with strings (As proofed by redis errors) | 19:18 |
| bnemec | Right | 19:19 |
| kmalloc | also python-memcached is effectively abandonware | 19:19 |
| kmalloc | 1 year since the last merge of code | 19:19 |
| kmalloc | and it is ... horribly designed. | 19:19 |
| dhellmann | efried : that looks reasonable to me. did you test it in the nova tests locally? | 19:19 |
| bnemec | Hmm, so pylibmc is the preferred way to talk to memcached? | 19:19 |
| kmalloc | and is explciitly thread-local focused | 19:19 |
| kmalloc | bnemec: i would use pymemcached | 19:19 |
| kmalloc | pymemcache* | 19:19 |
| kmalloc | but dogpile doesn't have that backend yet | 19:20 |
| efried | dhellmann: no, left a note asking melwitt to do that, but I can if she doesn't wanna. | 19:20 |
| kmalloc | so, maybe the solution is 1) fix python-memcached backend (in oslo.cache) | 19:20 |
| kmalloc | just a very basic fix. | 19:20 |
| kmalloc | and 2) write a backend that uses pymemcache | 19:20 |
| kmalloc | which is *very* good | 19:20 |
| dhellmann | efried : Ok. I think we want that class to show up somewhere in the docs. I don't remember if oslo.service uses the auto-doc stuff or not | 19:21 |
| kmalloc | i was going to do it all at once, but this seems like i have other issues such as serialization/deserialization | 19:21 |
| kmalloc | since i *dont* want to use pickle | 19:21 |
| dhellmann | efried : it looks like it uses it partially | 19:21 |
| kmalloc | unfortunately, complex objects seem to have issues with msgpack sometimes. | 19:21 |
| efried | dhellmann: Roger that. It's mostly PoC at this stage. LMK what needs to happen to make it goable. | 19:21 |
| dhellmann | efried : I just left a note on the review | 19:22 |
| dhellmann | thanks for working on this! | 19:22 |
| efried | dhellmann: ack, thanks | 19:22 |
| kmalloc | also, loading unverified data via pickle... can expose security issues | 19:22 |
| kmalloc | e.g. if the memcache server is compromised... the runtime code could be changed | 19:22 |
| kmalloc | msgpack is a bit better on that front. | 19:22 |
| dhellmann | efried, bnemec : should we do a reno for the fixture, too? or are we keeping those for production users? | 19:22 |
| bnemec | Yeah, I thought that was the main reason pickle is discouraged. | 19:22 |
| *** cdent has quit IRC | 19:23 | |
| kmalloc | bleh... this is icky. | 19:23 |
| kmalloc | so 1) i'll monkeypatch a fix for oslo.cache for python-memcache | 19:23 |
| bnemec | dhellmann: I think in the past you had said reno should be for user-facing things. But oslo is kind of weird in that in some cases our users are developers. | 19:23 |
| kmalloc | i'll make a new backend that isn't crummy and isn't using pickle | 19:23 |
| dhellmann | bnemec : yeah, I always means "production user" in that context, but I don't know if that's how others interpret it | 19:24 |
| kmalloc | and we can do a deprecation cycle where oslo.cache stops supporting python-memcache | 19:24 |
| kmalloc | and we allow folks to register the serializer/deserializer mechanisms/specs for msgpack where it breaks | 19:24 |
| kmalloc | mostly, in keysotne we have some wonky almost self-referential objects | 19:25 |
| kmalloc | which is bad in itself. | 19:25 |
| bnemec | dhellmann: Maybe we should add a developer notes section to reno? | 19:25 |
| bnemec | kmalloc: Yeah, that sounds like something we should do. | 19:25 |
| kmalloc | bnemec: hm i actually might be able to fix without a revert now that i think about it | 19:26 |
| melwitt | efried: how to try it locally? patch it into oslo.service in my unit test venv? | 19:26 |
| kmalloc | bnemec: if i have a fix that works sans revert, we'll roll with that before pushing the revert through | 19:26 |
| efried | melwitt: Yes, that should work. I usually git clone the project under .tox/pyXX/src and pip install -e it from within the venv; then I can do things like git review -d XXXXX to pull down different patch sets or whatever. | 19:27 |
| bnemec | kmalloc: Works for me. I already submitted the patch to block this release in g-r so as long as we don't release again nobody should be blocked. | 19:27 |
| melwitt | efried: kewl, thanks. I'm a noob on that | 19:28 |
| efried | melwitt: lmk if you need help | 19:28 |
| kmalloc | ++ | 19:28 |
| kmalloc | bnemec: perfect | 19:28 |
| * bnemec needs to go vote and eat | 19:30 | |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache master: Only split url arg with memcache backend https://review.openstack.org/615979 | 19:31 |
| kmalloc | bnemec: ^ that should be the "real" fix. | 19:32 |
| kmalloc | ooh or whoops, wont fix pool. | 19:33 |
| efried | dhellmann: Save me digging, since I reckon you'll know this one off the top: What's the sphinx/rst form for linking to another doc in the same tree? | 19:40 |
| openstackgerrit | Morgan Fainberg proposed openstack/oslo.cache master: Only split url arg with memcache backend https://review.openstack.org/615979 | 19:41 |
| efried | dhellmann: nm, figgered it out. | 19:42 |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Fixture to mock loopingcall wait() https://review.openstack.org/615978 | 19:46 |
| jroll | how does one set a StrOpt that has a default to None (the value, not the string)? is that possible? | 20:02 |
| jroll | I feel like it is, but cannot find how | 20:02 |
| efried | jroll: Is the default not None? | 20:08 |
| jroll | efried: correct | 20:10 |
| jroll | reading the code, it appears I'm wrong :( | 20:10 |
| *** bobh has joined #openstack-oslo | 20:11 | |
| efried | jroll: Meaning you got it figured out? | 20:12 |
| jroll | efried: meaning I believe that I cannot set the value to None using the config file, and would love to be wrong about that :) | 20:12 |
| efried | jroll: oic | 20:13 |
| efried | jroll: So you want a config opt that's either not None by default, and to set that guy to None via the config (why?), or you care about this because you want a mutable opt to be able to go from some value back to its default of None. | 20:13 |
| efried | jroll: In the latter case, does removing it from the conf do the trick? | 20:14 |
| jroll | efried: the former. see the help here: https://github.com/openstack/ironic/blob/master/ironic/conf/pxe.py#L35 | 20:14 |
| efried | hah | 20:14 |
| efried | jroll: What about using literally '<None>' | 20:15 |
| jroll | efried: does not work | 20:15 |
| efried | jroll: And you didn't even get lucky with the code doing an implicit falsey check. It's actually asking for `is None` :( | 20:18 |
| efried | jroll: I assume you've tried | 20:19 |
| efried | instance_master_path = | 20:19 |
| efried | and ended up with '' | 20:19 |
| jroll | efried: indeed | 20:19 |
| efried | jroll: Well, this has only been a problem since 2014. What's the big deal? | 20:20 |
| jroll | efried: :D | 20:20 |
| efried | jroll: Okay, a second pair of eyes has gone through both the ironic and oslo.config sides of this and confirmed yer scrood. Sorry. | 20:28 |
| efried | jroll: The simplest fix would seem to be changing the ironic side to check for '' as well as None, since that seems like it surely covers the intent. The empty string is not a sensible value for that option, right? | 20:31 |
| *** mmethot_ is now known as mmethot | 20:38 | |
| *** munimeha1 has quit IRC | 20:46 | |
| *** raildo has quit IRC | 21:06 | |
| jroll | efried: thanks for the extra eyes. agree that seems reasonable, I'll see what other ironic folks say :) | 21:08 |
| *** bobh has quit IRC | 21:20 | |
| *** pcaruana has quit IRC | 21:31 | |
| *** ansmith has quit IRC | 21:42 | |
| dhellmann | jroll : why do you want to set a string option to None? | 21:45 |
| *** bobh has joined #openstack-oslo | 21:46 | |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Limit monotonic to py2 https://review.openstack.org/616013 | 21:47 |
| efried | mordred: ^ | 21:47 |
| efried | dhellmann: https://github.com/openstack/ironic/blob/master/ironic/conf/pxe.py#L35 <== | 21:48 |
| efried | dhellmann: (to answer the question about string option to None) | 21:48 |
| efried | dhellmann: And we did check the rest of the ironic side code: it is actually doing 'is None' at the moment. Also the oslo code, which doesn't seem to provide a facility for explicitly setting to None. | 21:48 |
| dhellmann | it seems like None was a way to say "this is not set" and we can do that with the location information now | 21:49 |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Fixture to mock loopingcall wait() https://review.openstack.org/615978 | 21:49 |
| dhellmann | yeah, None isn't a valid string so I don't think I'd want to add some complex logic to let the user spell out None in some format in the config file | 21:49 |
| dhellmann | it just feels like doing extra work, right? | 21:49 |
| dhellmann | https://docs.openstack.org/oslo.config/latest/reference/locations.html feels like a better way if the question is "did the user set this value?" | 21:50 |
| efried | dhellmann: The consuming code seems crazy to me. I think any time you have a default that's a real string, it's reasonable not to make None a significant value. | 21:51 |
| dhellmann | I agree that it would be better to treat '' and None the same here | 21:51 |
| dhellmann | I think we weren't able to change the default of StringOpt because we have too many places looking for a None | 21:51 |
| efried | ++, I think that's what jroll is gonna jroll with. | 21:51 |
| efried | dhellmann: The default for StrOpt is indeed None. | 21:52 |
| efried | which is fine. | 21:52 |
| efried | It's just that this case sets a real default. | 21:52 |
| jroll | dhellmann: I did not know about the location information. that may work, I'll need to dig into it. thanks! | 21:52 |
| jroll | well, maybe not. but it's useful for other things :) | 21:53 |
| dhellmann | jroll : it's relatively new | 21:54 |
| dhellmann | last cycle I think? | 21:54 |
| jroll | neat. I like it. | 21:54 |
| *** geekinutah has quit IRC | 21:57 | |
| *** geekinutah has joined #openstack-oslo | 21:57 | |
| kmalloc | dhellmann: oslo.config simply ignores extra misc. data/options in config files right? | 21:58 |
| kmalloc | cc anyone ^ | 21:58 |
| dhellmann | kmalloc : yes | 21:58 |
| * kmalloc is unclear on following deprecation policy, specifically removal of options from, say, keystonemiddleware if all the functionality is removed. | 21:58 | |
| kmalloc | e.g. pki/pkiz code is long dead, keystone hasn't supported the token formats in... a long time | 21:59 |
| dhellmann | I think the validator will tell you if you have data that isn't attached to a registered option, but the production code ignores things in the file until the app tries to access the data | 21:59 |
| kmalloc | right. so, if we're removing all code that references the options... does it even make sense to keep them? | 21:59 |
| dhellmann | have the options been marked as deprecated? | 21:59 |
| kmalloc | no. | 21:59 |
| kmalloc | that is why i am asking | 21:59 |
| dhellmann | ok | 21:59 |
| kmalloc | the code is removed | 22:00 |
| kmalloc | and there is nothing that references it anywhere | 22:00 |
| kmalloc | zero impact on running service | 22:00 |
| dhellmann | the usual process is to mark the options as deprecated and then have the app emit warnings if the options are being set | 22:00 |
| kmalloc | bleh | 22:00 |
| dhellmann | but if the code is already gone and nothing uses them at all, you may not need to do that | 22:00 |
| kmalloc | right | 22:00 |
| kmalloc | well it isn't gone yet | 22:00 |
| kmalloc | but it is bitrotting and untested in any real way | 22:00 |
| dhellmann | hmm | 22:00 |
| efried | If you wanted to be really nice, you could say "these were removed in XXX release and are ignored" and then delete them in the next release. | 22:01 |
| kmalloc | the code could not be used at all for ... like 2 yrs | 22:01 |
| dhellmann | so it's there but doesn't work? | 22:01 |
| kmalloc | right | 22:01 |
| kmalloc | it can't work with modern keystones | 22:01 |
| dhellmann | is it the sort of thing someone might have worked around without upstreaming that fix? which isn't to say i care that much about those people, but just for completeness... | 22:01 |
| kmalloc | since keystone does not emit pki/pkiz tokens | 22:01 |
| kmalloc | unlikely. | 22:02 |
| *** bobh has quit IRC | 22:02 | |
| dhellmann | I think it's probably safe to just delete the stuff then | 22:02 |
| kmalloc | that was my view. | 22:02 |
| kmalloc | but i wanted to check and make sure i'm making the correct case for this | 22:02 |
| dhellmann | the advance warning would have been the exception thrown from the broken code | 22:02 |
| kmalloc | ++ | 22:02 |
| kmalloc | well keystonemiddleware wouldn't receive a pki token anymore | 22:03 |
| dhellmann | it may still end up triggering a major version bump in the next release of the lib, but that's a separate conversation | 22:03 |
| kmalloc | keystonemiddleware always gets a major bump, iirc or it used to | 22:03 |
| dhellmann | I'm not sure why it would always, but maybe it does | 22:03 |
| kmalloc | because it is tied to library versions and can't be mixed and matched. has to be run in the same processspace as <service> | 22:03 |
| kmalloc | it used to because it was versionlocked. | 22:04 |
| kmalloc | e.g. it needs version X of keystoneclient (for example) and therefore nova would also only ever be able to use that version of ksc | 22:04 |
| kmalloc | because it runs in the same processpace as nova. | 22:04 |
| kmalloc | it may no longer have that. | 22:04 |
| kmalloc | but at one point it was managed like that, same as a service. | 22:04 |
| kmalloc | dhellmann: *shrug*. | 22:06 |
| dhellmann | ah, ok | 22:07 |
| kmalloc | i think i can knock down the imports to KSA and a couple other minor things that should make it much less... obnoxious soon | 22:09 |
| *** kgiusti has left #openstack-oslo | 22:09 | |
| kmalloc | ideally, it would have oslo.config, keystoneauth and ... maybe webob? | 22:09 |
| kmalloc | i'll figure that out once a lot of the dead code is removed :) | 22:09 |
| *** dave-mccowan has quit IRC | 22:35 | |
| *** bobh has joined #openstack-oslo | 22:42 | |
| *** bobh has quit IRC | 22:45 | |
| *** bobh has joined #openstack-oslo | 22:46 | |
| *** bobh has quit IRC | 22:50 | |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Limit monotonic to py2 https://review.openstack.org/616013 | 23:42 |
| openstackgerrit | Eric Fried proposed openstack/oslo.service master: Fixture to mock loopingcall wait() https://review.openstack.org/615978 | 23:43 |
| efried | bnemec: Done, let's see if that flies... | 23:43 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!