| *** pm90_ has joined #openstack-oslo | 00:06 | |
| *** ndipanov has joined #openstack-oslo | 00:06 | |
| *** pm90_ has quit IRC | 00:14 | |
| *** achanda has joined #openstack-oslo | 00:16 | |
| *** jeckersb_gone is now known as jeckersb | 00:17 | |
| *** harlowja has joined #openstack-oslo | 00:20 | |
| *** achanda has quit IRC | 00:25 | |
| *** mtanino has quit IRC | 00:28 | |
| *** achanda has joined #openstack-oslo | 00:30 | |
| openstackgerrit | Joshua Harlow proposed openstack/tooz: Add programatic introspection of drivers capabilities https://review.openstack.org/240681 | 00:32 |
|---|---|---|
| openstackgerrit | Joshua Harlow proposed openstack/tooz: Add decorator that can be used to ensure a driver is capable https://review.openstack.org/241056 | 00:33 |
| *** dims_ has quit IRC | 00:33 | |
| *** rohit_ has joined #openstack-oslo | 00:33 | |
| *** jamielennox is now known as jamielennox|away | 00:33 | |
| *** dims has joined #openstack-oslo | 00:36 | |
| *** sputnik13 has quit IRC | 00:45 | |
| *** dims_ has joined #openstack-oslo | 00:51 | |
| *** dims has quit IRC | 00:51 | |
| *** _amrith_ is now known as amrith | 00:52 | |
| harlowja | dims_ https://gist.github.com/harlowja/099cf904365d7fed31e7 was what i am starting to form to maybe get some new core recruits | 01:10 |
| harlowja | mission impossible themed ^ ;) | 01:10 |
| harlowja | run like | 01:11 |
| harlowja | $ python maybe_oslo_core.py "John" "Doe" "oslo.messaging" | 01:11 |
| dims_ | harlowja nice! got a list of people too? | 01:11 |
| harlowja | can do, gotta analyze that git-inspector report | 01:11 |
| *** rjaiswal has quit IRC | 01:15 | |
| *** thumpba has joined #openstack-oslo | 01:16 | |
| *** thumpba has quit IRC | 01:18 | |
| *** achanda has quit IRC | 01:29 | |
| *** SurajD has joined #openstack-oslo | 01:32 | |
| harlowja | SpamapS ok, i referenced your tooz policy in https://review.openstack.org/#/c/209661/ | 01:37 |
| harlowja | denoted that your tooz policy spec will be the thing that decided if a tooz driver lives/dies (and/or what it must conform to) | 01:37 |
| *** amrith is now known as _amrith_ | 01:45 | |
| *** _amrith_ is now known as amrith | 01:46 | |
| *** yamamoto has joined #openstack-oslo | 01:53 | |
| openstackgerrit | Joshua Harlow proposed openstack/oslo.service: Avoid the dual-naming confusion https://review.openstack.org/238717 | 01:54 |
| openstackgerrit | Suraj Deshmukh proposed openstack/oslo.utils: Added ICMP 'type' and 'code' checking capability to 'netutils' module https://review.openstack.org/240661 | 01:54 |
| *** mc_nair has quit IRC | 01:54 | |
| SurajD | Can somebody review this https://review.openstack.org/#/c/240661/ | 01:55 |
| *** pm90_ has joined #openstack-oslo | 02:00 | |
| openstackgerrit | Joshua Harlow proposed openstack/oslo.utils: Add useful 'time_it' decorator https://review.openstack.org/231220 | 02:00 |
| openstackgerrit | Davanum Srinivas (dims) proposed openstack/oslo.messaging: Decouple transport for RPC and Notification https://review.openstack.org/233258 | 02:01 |
| *** dims_ has quit IRC | 02:01 | |
| *** dims has joined #openstack-oslo | 02:05 | |
| *** SurajD has quit IRC | 02:05 | |
| openstackgerrit | Merged openstack/tooz: Updated from global requirements https://review.openstack.org/239048 | 02:06 |
| *** SurajD has joined #openstack-oslo | 02:08 | |
| *** bana_k has quit IRC | 02:10 | |
| *** salv-orlando has joined #openstack-oslo | 02:12 | |
| *** yamamoto has quit IRC | 02:14 | |
| *** yamamoto has joined #openstack-oslo | 02:17 | |
| *** yamamoto has quit IRC | 02:21 | |
| *** SurajD has quit IRC | 02:22 | |
| *** ozamiatin has joined #openstack-oslo | 02:29 | |
| *** jamielennox|away is now known as jamielennox | 02:34 | |
| *** ozamiatin has quit IRC | 02:37 | |
| *** pm90_ has quit IRC | 02:42 | |
| *** rohit_ has quit IRC | 02:46 | |
| *** bapalm has quit IRC | 02:54 | |
| openstackgerrit | Merged openstack/oslo-specs: Add spec for tooz driver policy https://review.openstack.org/240373 | 02:54 |
| *** thumpba has joined #openstack-oslo | 02:56 | |
| *** thumpba has quit IRC | 03:02 | |
| *** rjaiswal has joined #openstack-oslo | 03:02 | |
| *** bapalm has joined #openstack-oslo | 03:02 | |
| *** yamamoto has joined #openstack-oslo | 03:07 | |
| *** dims has quit IRC | 03:13 | |
| rjaiswal | dims: https://review.openstack.org/#/c/243360/ for bumping up the requirements | 03:16 |
| openstackgerrit | Merged openstack/oslo.versionedobjects: Ensure__repr__ return value is encoded https://review.openstack.org/242964 | 03:41 |
| openstackgerrit | Merged openstack/oslo.messaging: Move supported messaging drivers in-tree https://review.openstack.org/240608 | 03:49 |
| openstackgerrit | Merged openstack/oslo.privsep: Initial basic privsep functionality https://review.openstack.org/238333 | 03:58 |
| *** links has joined #openstack-oslo | 03:59 | |
| *** salv-orlando has quit IRC | 03:59 | |
| *** mriedem_away has quit IRC | 04:05 | |
| *** david-lyle has joined #openstack-oslo | 04:07 | |
| *** thumpba has joined #openstack-oslo | 04:22 | |
| *** thumpba has quit IRC | 04:26 | |
| *** jerrygb has quit IRC | 04:28 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/oslo.privsep: Updated from global requirements https://review.openstack.org/243405 | 04:33 |
| *** jerrygb has joined #openstack-oslo | 04:36 | |
| *** amotoki has joined #openstack-oslo | 04:50 | |
| *** salv-orlando has joined #openstack-oslo | 05:01 | |
| *** jamielennox is now known as jamielennox|away | 05:01 | |
| *** jamespage has quit IRC | 05:05 | |
| *** salv-orlando has quit IRC | 05:05 | |
| *** jamespage has joined #openstack-oslo | 05:06 | |
| *** yamamoto_ has joined #openstack-oslo | 05:13 | |
| *** yamamoto has quit IRC | 05:15 | |
| *** jerrygb has quit IRC | 05:20 | |
| *** gcb has joined #openstack-oslo | 05:22 | |
| *** jerrygb has joined #openstack-oslo | 05:23 | |
| *** jerrygb has quit IRC | 05:23 | |
| openstackgerrit | Akihiro Motoki proposed openstack/oslo.config: Add maxlen check to StrOpt https://review.openstack.org/242880 | 05:49 |
| *** ozamiatin has joined #openstack-oslo | 05:51 | |
| *** achanda has joined #openstack-oslo | 06:03 | |
| *** ig0r__ has quit IRC | 06:05 | |
| *** ig0r__ has joined #openstack-oslo | 06:10 | |
| *** bana_k has joined #openstack-oslo | 06:11 | |
| *** jerrygb has joined #openstack-oslo | 06:24 | |
| *** thumpba has joined #openstack-oslo | 06:24 | |
| *** jerrygb has quit IRC | 06:28 | |
| *** thumpba has quit IRC | 06:29 | |
| *** rjaiswal has quit IRC | 06:35 | |
| *** itisha has joined #openstack-oslo | 06:57 | |
| *** alexpilotti has joined #openstack-oslo | 07:15 | |
| *** nkrinner has joined #openstack-oslo | 07:20 | |
| *** bana_k has quit IRC | 07:21 | |
| *** jerrygb has joined #openstack-oslo | 07:25 | |
| *** thumpba has joined #openstack-oslo | 07:25 | |
| *** achanda has quit IRC | 07:27 | |
| *** jerrygb has quit IRC | 07:29 | |
| *** thumpba has quit IRC | 07:29 | |
| *** alexpilotti has quit IRC | 07:35 | |
| *** bana_k has joined #openstack-oslo | 07:37 | |
| *** openstackgerrit has quit IRC | 07:46 | |
| *** openstackgerrit has joined #openstack-oslo | 07:46 | |
| *** ozamiatin_ has joined #openstack-oslo | 07:48 | |
| *** ozamiatin has quit IRC | 07:51 | |
| *** zigo has quit IRC | 08:01 | |
| *** zigo has joined #openstack-oslo | 08:03 | |
| *** ozamiatin_ has quit IRC | 08:06 | |
| *** bana_k has quit IRC | 08:10 | |
| *** ozamiatin_ has joined #openstack-oslo | 08:22 | |
| *** browne has quit IRC | 08:22 | |
| *** ozamiatin_ has quit IRC | 08:22 | |
| *** jerrygb has joined #openstack-oslo | 08:26 | |
| *** jerrygb has quit IRC | 08:31 | |
| *** openstackgerrit has quit IRC | 08:31 | |
| *** openstackgerrit has joined #openstack-oslo | 08:31 | |
| *** ozamiatin_ has joined #openstack-oslo | 08:35 | |
| *** ozamiatin_ has quit IRC | 08:46 | |
| *** ozamiatin_ has joined #openstack-oslo | 08:56 | |
| *** achanda has joined #openstack-oslo | 08:57 | |
| *** achanda has quit IRC | 09:02 | |
| *** yassine has joined #openstack-oslo | 09:23 | |
| *** jerrygb has joined #openstack-oslo | 09:27 | |
| *** jerrygb has quit IRC | 09:32 | |
| *** ihrachys has joined #openstack-oslo | 09:35 | |
| *** achanda has joined #openstack-oslo | 10:00 | |
| *** achanda has quit IRC | 10:05 | |
| openstackgerrit | Vladimir Eremin proposed openstack/oslo.config: [WIP] Support ZooKeeper and Consul for oslo.config https://review.openstack.org/243182 | 10:08 |
| *** ozamiatin_ has quit IRC | 10:08 | |
| *** gcb has quit IRC | 10:09 | |
| openstackgerrit | Merged openstack/oslo.service: Forbid launching services with 0 or negative number of workers https://review.openstack.org/241240 | 10:22 |
| *** ozamiatin has joined #openstack-oslo | 10:22 | |
| haypo | i found 11 clients using cliutils.py (of oslo-incubator). on these 11 clients, 8 uses oslo.utils, but 3 (mistraclient, solumclient, tuskarclient) don't use oslo.utils yet | 10:22 |
| haypo | would it make sense to graduate cliutils.py to oslo.utils? | 10:22 |
| *** jamielennox|away is now known as jamielennox | 10:24 | |
| *** jerrygb has joined #openstack-oslo | 10:28 | |
| *** nikhil_k has joined #openstack-oslo | 10:29 | |
| *** nikhil has quit IRC | 10:30 | |
| *** jerrygb has quit IRC | 10:34 | |
| *** cdent has joined #openstack-oslo | 10:35 | |
| *** ozamiatin has quit IRC | 10:37 | |
| *** ihrachys has quit IRC | 10:38 | |
| *** ihrachys has joined #openstack-oslo | 10:39 | |
| *** pblaho has quit IRC | 10:39 | |
| *** ig0r__ has quit IRC | 10:40 | |
| *** yamamoto_ has quit IRC | 10:44 | |
| haypo | jd__: the openstack-sdk is https://github.com/openstack/python-openstacksdk ? | 11:02 |
| jd__ | haypo: yes | 11:02 |
| haypo | jd__: if i understood correctly, openstacksdk is a new implementation of many clients. but i don't understand what we are supposed to do with novaclient | 11:04 |
| haypo | novaclient should be a thin wrapper to openstacksdk? | 11:04 |
| jd__ | haypo: maybe? how would I know | 11:04 |
| haypo | jd__: what is your plan for ceilometer? | 11:04 |
| jd__ | we don't really have plan for ceilometer since most of its api is moving away | 11:05 |
| jd__ | but for gnocchiclient and aodhclient we don't plan on working with openstack-sdk | 11:05 |
| *** dims has joined #openstack-oslo | 11:05 | |
| jd__ | we just use keystoneauth1 | 11:05 |
| *** ozamiatin has joined #openstack-oslo | 11:07 | |
| *** ihrachys has quit IRC | 11:18 | |
| openstackgerrit | Julien Danjou proposed openstack/oslo.messaging: Revert "Robustify locking in MessageHandlingServer" https://review.openstack.org/243538 | 11:19 |
| *** ihrachys has joined #openstack-oslo | 11:19 | |
| *** dims has quit IRC | 11:21 | |
| haypo | jd__: do you know where does the "openstack" CLI command come from? | 11:21 |
| *** itisha has quit IRC | 11:21 | |
| jd__ | haypo: openstackclient | 11:21 |
| *** dims has joined #openstack-oslo | 11:22 | |
| haypo | hum, and openstackclient doesn't use openstacksdk | 11:23 |
| *** ihrachys has quit IRC | 11:23 | |
| *** yamamoto has joined #openstack-oslo | 11:35 | |
| *** yamamoto has quit IRC | 11:39 | |
| *** yamamoto has joined #openstack-oslo | 11:39 | |
| *** yamamoto_ has joined #openstack-oslo | 11:40 | |
| *** jaypipes has joined #openstack-oslo | 11:42 | |
| *** yamamoto has quit IRC | 11:43 | |
| *** ihrachys has joined #openstack-oslo | 11:49 | |
| *** achanda has joined #openstack-oslo | 11:54 | |
| *** yamamoto_ has quit IRC | 12:06 | |
| *** ajo is now known as ajo\lunch | 12:14 | |
| akwasnie_ | hey, i am trying to configure oslo log to log tracebacks to syslog, but it seems that it does not work form me. I have set syslog_level to debug and use_syslog to true, but still there is no sign of any tracebacks in syslog. errors and infos are logged properly | 12:18 |
| akwasnie_ | any guide on how to fix this? | 12:21 |
| *** gordc has joined #openstack-oslo | 12:22 | |
| *** achanda has quit IRC | 12:23 | |
| dims | haypo: at this point i am thinking of oslo.cli = apiclient/*.py + cliutils.py | 12:25 |
| *** freyes has quit IRC | 12:26 | |
| *** freyes has joined #openstack-oslo | 12:26 | |
| *** yamamoto has joined #openstack-oslo | 12:27 | |
| *** jerrygb has joined #openstack-oslo | 12:32 | |
| *** jerrygb has quit IRC | 12:37 | |
| *** dims has quit IRC | 12:53 | |
| *** amrith is now known as _amrith_ | 12:54 | |
| *** dims has joined #openstack-oslo | 12:56 | |
| *** ajo\lunch is now known as ajo | 12:57 | |
| openstackgerrit | Daisuke Fujita proposed openstack/oslo.log: The user_identity format flexibility https://review.openstack.org/218139 | 12:58 |
| *** pblaho has joined #openstack-oslo | 13:03 | |
| *** ozamiatin has quit IRC | 13:04 | |
| haypo | dims: ah yes i was also thinking at puting cliutils inside apiclient | 13:08 |
| openstackgerrit | Markus Zoeller (markus_z) proposed openstack/oslo.config: Add help text generation for OptGroups https://review.openstack.org/242472 | 13:09 |
| haypo | dims: that's also why i read the spec and the mail about apiclient ;) | 13:09 |
| *** stevemar_ has joined #openstack-oslo | 13:11 | |
| openstackgerrit | Julien Danjou proposed openstack/oslo-incubator: Remove deprecated `apiclient' https://review.openstack.org/243578 | 13:13 |
| openstackgerrit | Julien Danjou proposed openstack/oslo-incubator: tests: remove unused testmods https://review.openstack.org/243579 | 13:16 |
| *** gcb has joined #openstack-oslo | 13:17 | |
| *** stevemar_ has quit IRC | 13:20 | |
| *** stevemar_ has joined #openstack-oslo | 13:21 | |
| *** achanda has joined #openstack-oslo | 13:23 | |
| *** salv-orlando has joined #openstack-oslo | 13:23 | |
| *** jerrygb has joined #openstack-oslo | 13:24 | |
| *** fultonj has joined #openstack-oslo | 13:24 | |
| *** yamamoto has quit IRC | 13:27 | |
| *** jerrygb has quit IRC | 13:29 | |
| *** ozamiatin has joined #openstack-oslo | 13:30 | |
| *** achanda has quit IRC | 13:30 | |
| haypo | dims, jd__ : apiclient/, cliutils.py, cliff, keystoneauth, openstackclient, openstacksdk, ... ok, now i'm lost :-) | 13:32 |
| haypo | but we probably have too many choices, creating yet another library may not help :) | 13:33 |
| *** kgiusti has joined #openstack-oslo | 13:34 | |
| jd__ | that's why I'm proposing to remove apiclient | 13:35 |
| haypo | jd__: ok, i replied on the list. i found projects using cliff *and* cliutils. there is maybe something that we can do to remove cliutils.py from clients too? | 13:53 |
| openstackgerrit | ChangBo Guo(gcb) proposed openstack/oslo.config: Allow method set_override with parameter override=None for all types https://review.openstack.org/243001 | 13:53 |
| jd__ | haypo: I would think so | 13:54 |
| haypo | jd__: i'm not sure that i understood the whole story, but i like the idea of having a single obvious way to write clients :) | 13:54 |
| haypo | and i understand that existing clients may use the "old" way ;) | 13:55 |
| *** yamamoto has joined #openstack-oslo | 13:56 | |
| *** rlrossit has joined #openstack-oslo | 13:59 | |
| *** stevemar_ has quit IRC | 14:04 | |
| *** links has quit IRC | 14:08 | |
| *** thumpba has joined #openstack-oslo | 14:09 | |
| rbradfor | I got a jenkins failure on review in a gate-tempest-dsvm-neutron-src-oslo.i18n gate, however the change is only in docs. Is this a gate that sometimes fails | 14:13 |
| *** stevemar_ has joined #openstack-oslo | 14:13 | |
| haypo | rbradfor: yes :-( | 14:14 |
| haypo | rbradfor: most gates are unstable | 14:14 |
| haypo | rbradfor: (functional and integration tests) | 14:14 |
| rbradfor | haypo, thanks. I'll just resubmit then. | 14:15 |
| haypo | rbradfor: just add the comment "recheck" | 14:15 |
| haypo | rbradfor: it schedules a recheck of all tests | 14:15 |
| rbradfor | haypo, that's even better. I was unaware of that tip. thanks. | 14:15 |
| haypo | rbradfor: see my notes, http://haypo-notes.readthedocs.org/openstack.html#hack-openstack | 14:17 |
| openstackgerrit | Vladimir Eremin proposed openstack/oslo.config: [WIP] Support ZooKeeper and Consul for oslo.config https://review.openstack.org/243182 | 14:18 |
| rbradfor | haypo, thanks | 14:24 |
| *** ndipanov has quit IRC | 14:24 | |
| *** yamamoto has quit IRC | 14:27 | |
| *** freyes has quit IRC | 14:30 | |
| *** _amrith_ is now known as amrith | 14:33 | |
| *** pballand has quit IRC | 14:35 | |
| haypo | dims, jd__, sileht : cool enhancement for oslo.config to review, https://review.openstack.org/#/c/221394/ | 14:37 |
| *** yassine has quit IRC | 14:39 | |
| *** regXboi has joined #openstack-oslo | 14:40 | |
| *** takedakn has joined #openstack-oslo | 14:40 | |
| dims | haypo +2A! | 14:42 |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:45 | |
| *** yassine has joined #openstack-oslo | 14:45 | |
| *** salv-orlando has quit IRC | 14:46 | |
| openstackgerrit | Victor Stinner proposed openstack/oslo.config: Remove "Kept for backward-compatibility" comment https://review.openstack.org/243630 | 14:46 |
| *** ujjain has quit IRC | 14:46 | |
| *** thangp has joined #openstack-oslo | 14:46 | |
| *** pradk has joined #openstack-oslo | 14:47 | |
| *** ujjain has joined #openstack-oslo | 14:48 | |
| *** ujjain has quit IRC | 14:48 | |
| *** ujjain has joined #openstack-oslo | 14:48 | |
| gcb | haypo, can you help review a python 3 issue fix: https://review.openstack.org/#/c/241591/ | 14:49 |
| *** mriedem has joined #openstack-oslo | 14:51 | |
| *** freyes has joined #openstack-oslo | 14:53 | |
| haypo | gcb: i'm not sure that "nested" is the right name | 14:53 |
| haypo | gcb: it's more an issue with the qualified name, a new feature of 3.3 or 3.4 (i don't recall) | 14:53 |
| *** stevemar_ has quit IRC | 14:53 | |
| haypo | nested class is more something like: class A: class B: pass | 14:53 |
| openstackgerrit | Merged openstack/oslo.middleware: add missing shortcut for HTTPProxyToWSGI middleware https://review.openstack.org/243212 | 14:54 |
| *** stevemar_ has joined #openstack-oslo | 14:54 | |
| gcb | haypo, the description may be not correct, I met the problem when port oslo.utils.reflection to keyston in https://review.openstack.org/#/c/241494/ | 14:55 |
| *** cprmrf has joined #openstack-oslo | 14:57 | |
| haypo | gcb: hum, the docstring says that qualified name is used | 14:58 |
| haypo | gcb: and there is an optional fully_qualified parameter | 14:58 |
| gcb | parameter fully_qualified doesn't handle the case | 14:58 |
| haypo | gcb: IMHO you must use __qualname__ if fully_qualified, or __name__ otherwise | 14:58 |
| *** fultonj has left #openstack-oslo | 14:59 | |
| haypo | gcb: use __name__ instead of using .rsplit('.')[-1], it's simpler ;) | 14:59 |
| *** mc_nair has joined #openstack-oslo | 14:59 | |
| *** amotoki_ has joined #openstack-oslo | 15:00 | |
| *** links has joined #openstack-oslo | 15:01 | |
| *** cprmrf has quit IRC | 15:03 | |
| haypo | dims: thanks for approving the oslo.config change ;) | 15:03 |
| *** links has quit IRC | 15:05 | |
| *** cprmrf has joined #openstack-oslo | 15:06 | |
| *** cprmrf__ has joined #openstack-oslo | 15:11 | |
| *** takedakn has quit IRC | 15:12 | |
| *** pradk has quit IRC | 15:13 | |
| *** pradk_ has joined #openstack-oslo | 15:13 | |
| gcb | haypo, in Python 3 obj.__qualname__ returns 'ClassA.__init__.<locals>.ClassB', In python2 obj has no attribue __qualname__, so back to obj.__name__ returns 'ClassB', it returns different name. That 's the problem. Do you mean only use obj.__name__ ? | 15:14 |
| *** pradk_ is now known as pradk | 15:14 | |
| gcb | you can get test test code in https://bugs.launchpad.net/oslo.utils/+bug/1513067 | 15:14 |
| openstack | Launchpad bug 1513067 in oslo.utils "reflection: need clean nested class's name with prefix in Python 3" [Undecided,In progress] - Assigned to ChangBo Guo(gcb) (glongwave) | 15:14 |
| *** cprmrf has quit IRC | 15:15 | |
| gcb | haypo, I saw your comments thanks | 15:17 |
| *** pm90_ has joined #openstack-oslo | 15:18 | |
| *** harlowja_at_home has joined #openstack-oslo | 15:20 | |
| *** pm90__ has joined #openstack-oslo | 15:24 | |
| *** rlrossit has quit IRC | 15:25 | |
| *** rlrossit has joined #openstack-oslo | 15:25 | |
| *** pm90_ has quit IRC | 15:26 | |
| *** achanda has joined #openstack-oslo | 15:27 | |
| *** yamamoto has joined #openstack-oslo | 15:28 | |
| openstackgerrit | ChangBo Guo(gcb) proposed openstack/oslo.utils: Fix obj.__qualname__ return class name with prefix in python 3 https://review.openstack.org/241591 | 15:28 |
| rbradfor | dims, from our conversation yesterday on _i18n -- https://review.openstack.org/#/c/243321/. | 15:31 |
| *** rjaiswal has joined #openstack-oslo | 15:31 | |
| rbradfor | This lead me to seeing how much more documentation is needed to create working examples, e.g. http://paste.openstack.org/show/478443/ | 15:31 |
| rjaiswal | 7:16 PM <rjaiswal> dims: https://review.openstack.org/#/c/243360/ for bumping up the requirements | 15:32 |
| *** achanda has quit IRC | 15:32 | |
| *** jerrygb has joined #openstack-oslo | 15:33 | |
| *** yamamoto has quit IRC | 15:33 | |
| rbradfor | I saw there was a Tokyo session on Better documentation. Has anybody thought about the approach to blend documentation with working demos, in a documenting plus cut/paste situation | 15:33 |
| openstackgerrit | Sean Dague proposed openstack/oslo.messaging: Revert "Robustify locking in MessageHandlingServer" https://review.openstack.org/243658 | 15:33 |
| sdague | dims: ^^^ revert proposed | 15:33 |
| *** jecarey has joined #openstack-oslo | 15:34 | |
| openstackgerrit | Sean Dague proposed openstack/oslo.messaging: Revert "Robustify locking in MessageHandlingServer" https://review.openstack.org/243658 | 15:34 |
| dims | sdague : +2A | 15:34 |
| *** gcb has quit IRC | 15:35 | |
| mriedem | dims: looks like that waiting log message is hitting some ceilometer/aodh tests too http://logstash.openstack.org/#dashboard/file/logstash.json?query=message:%5C%22has%20been%20waiting%20for%5C%22 | 15:36 |
| mriedem | gordc: ^ | 15:36 |
| *** jerrygb has quit IRC | 15:37 | |
| dims | mdbooth harlowja ping about oslo.messaging issues | 15:37 |
| gordc | mriedem: dims: not sure if this is root of issue but we do have this: https://review.openstack.org/#/c/243538/ | 15:38 |
| dims | gordc i approved sdague's revert for the same | 15:39 |
| gordc | dims: kk, i guess jd__ can drop his. | 15:40 |
| jd__ | whaat | 15:42 |
| jd__ | did sdague just stole my patch | 15:42 |
| dims | jd__ haha :) | 15:42 |
| gordc | patch jacked | 15:43 |
| sdague | jd__: sorry, we didn't see it when doing debug this morning | 15:43 |
| *** salv-orlando has joined #openstack-oslo | 15:46 | |
| jd__ | sdague: no offense lol | 15:46 |
| harlowja_at_home | hmmmm, locking issue | 15:52 |
| harlowja_at_home | i'm pretty sure if oslo.messaging was written in go, using redis, it would be all better | 15:53 |
| harlowja_at_home | (snark snark) | 15:53 |
| dims | LOL | 15:53 |
| dims | no redis, we need consul or etcd | 15:53 |
| harlowja_at_home | reconsuld | 15:54 |
| harlowja_at_home | we need ^ | 15:54 |
| harlowja_at_home | (i just made that up) | 15:54 |
| *** freyes has quit IRC | 15:54 | |
| harlowja_at_home | hmmm, i wonder what happened with the locking | 15:55 |
| harlowja_at_home | guess we must of missed some logic | 15:55 |
| * harlowja_at_home will ponder | 15:55 | |
| sdague | harlowja_at_home: I think it did not anticipate the eventlet case | 15:56 |
| *** salv-orlando has quit IRC | 15:57 | |
| *** freyes has joined #openstack-oslo | 15:57 | |
| *** mtanino has joined #openstack-oslo | 15:57 | |
| harlowja_at_home | sdague, do u mean not using green conditions or something else? | 15:58 |
| harlowja_at_home | (if u know/have ideas) | 15:58 |
| sdague | harlowja_at_home: don't know entirely, but nova, for instance, uses the fakeimpl driver to build an entire set of nova services in a single process (including possibly multiple computes) | 15:59 |
| sdague | and run scenarios just like a real environment | 15:59 |
| sdague | so we're pushing back and forth on that a bunch | 15:59 |
| sdague | all with greenlets | 15:59 |
| harlowja_at_home | agreed, so probably something eventlet not being tweaked right | 16:00 |
| harlowja_at_home | monkey patching threading in that scenario happens though right? | 16:00 |
| harlowja_at_home | *pretty sure in general nova does that | 16:00 |
| harlowja_at_home | anyways, we'll figure it out | 16:03 |
| *** jecarey has quit IRC | 16:06 | |
| *** jerrygb has joined #openstack-oslo | 16:07 | |
| * harlowja_at_home pretty sure we just need to simplify '_OrderedTask' in that revert, make it a little more explict whats going on, and in smaller chunks, that will probably help | 16:08 | |
| *** jerrygb has quit IRC | 16:12 | |
| kevinbenton | zzzeek: ping re sqlalchemy and polymorphic_identity. got a quick question | 16:13 |
| sdague | anyway, the nova unit tests definitely expose this 100% locally, so you can use that as a test on this to figure out if it's right | 16:13 |
| *** jerrygb has joined #openstack-oslo | 16:13 | |
| zzzeek | kevinbenton: sure | 16:13 |
| kevinbenton | zzzeek: so in the Employee Manager example, say the employee ID is an auto increment value and that's how manager is related to employee | 16:14 |
| kevinbenton | zzzeek: does sqlalchemy handle the legwork of inserting a manager record of inserting an employee record and using the auto inc value? | 16:14 |
| *** jerrygb_ has joined #openstack-oslo | 16:15 | |
| zzzeek | kevinbenton: yes | 16:15 |
| *** pballand has joined #openstack-oslo | 16:15 | |
| kevinbenton | zzzeek: excellent, any hints i need to give it or is it good to go? | 16:15 |
| harlowja_at_home | sdague, great, that will def help | 16:16 |
| kevinbenton | zzzeek: other than 'polymorphic_identity' of course | 16:16 |
| zzzeek | kevinbenton: in joined table inheritance the child and parent tables are related to each other using a foreign key as usual. the mapper detects the column in the chlid table that is to be "synchronized" with the parent primary key column. If not foreign keys, this relationship can also be configured using the "inherit_condition" mapper argument. | 16:17 |
| zzzeek | kevinbenton: the mapper runs an INSERT for the parent table first, fetches the autoinc PK, and "synchronizes" it into the child table value | 16:17 |
| *** jerrygb has quit IRC | 16:17 | |
| kevinbenton | zzzeek: very cool. this is in the context of that neutron patch i was working on before so when i upload a new version you should get a gerrit notice | 16:18 |
| *** amotoki_ has quit IRC | 16:19 | |
| zzzeek | kevinbenton: sure | 16:20 |
| kevinbenton | zzzeek: thanks for your time! | 16:20 |
| zzzeek | kevinbenton: i hope the joined inh is really waht we need there, joined inh is pricey | 16:21 |
| kevinbenton | zzzeek: well based on jay's comments, it sounds like it shouldn't be a big hit if we are using ints | 16:22 |
| kevinbenton | zzzeek: i will do some profiling though to verify | 16:22 |
| zzzeek | kevinbenton: the joins...are heavy. they are much less heavy in SQLA 1.0 though, we used to rely on subqueries a lot which completely blew mysql away | 16:22 |
| *** jecarey has joined #openstack-oslo | 16:23 | |
| *** ozamiatin has quit IRC | 16:23 | |
| kevinbenton | zzzeek: oh, i see. you mean the way that sqlalchemy is rendering the queries | 16:23 |
| mdbooth | harlowja dims: Ooh, fun | 16:24 |
| mdbooth | Example? | 16:24 |
| zzzeek | kevinbenton: well yes, the classic case is loading Foo objects that have a collection of Manager, the query needs a right-nested join. sqlite doesnt support this and maybe mysql didnt a long time ago so we would turn the right-nest into a subuqery | 16:24 |
| zzzeek | select foo.*, subq.* from foo join (select * from employee join employee on <onclause>) on <onclause> | 16:24 |
| *** nkrinner has quit IRC | 16:25 | |
| zzzeek | ^^^^ brings mysql to a halt | 16:25 |
| harlowja_at_home | mdbooth, sdague should be able to point u at the way to run the tests he's talking about, supposedly it always happens, so thats good | 16:25 |
| harlowja_at_home | i'm wondering if its the ordering stuffs | 16:25 |
| mdbooth | dims: I'd be -1 on the revert without seeing an example | 16:25 |
| harlowja_at_home | perhaps something off there... | 16:25 |
| zzzeek | in 1.0 we do: select foo.*, employee.*, manager.* from foo join (employee join manager on <onclause>) on <onclause> | 16:25 |
| mdbooth | The usage of eventlet in the tests shouldn't be relevant. That basically just makes testing easier. | 16:25 |
| zzzeek | s/employee join employee/employee join manager/ | 16:25 |
| sdague | mdbooth: go look at the bug | 16:25 |
| dims | mdbooth check nova logs too please | 16:25 |
| kevinbenton | zzzeek: makes sense. is 1.0 in our requirements for openstack though? | 16:26 |
| dims | irc logs | 16:26 |
| zzzeek | kevinbenton: yes | 16:26 |
| mdbooth | sdague: I looked. Didn't see an example. | 16:26 |
| mdbooth | Perhaps I looked too quickly. | 16:26 |
| sdague | mdbooth: define "example" | 16:26 |
| mdbooth | Ah, ha | 16:26 |
| mdbooth | That's a different bug :) | 16:26 |
| sdague | the point is that tests randomly end up taking up to the max timeout | 16:26 |
| mdbooth | I was looking at a pretty useless bug | 16:26 |
| kevinbenton | zzzeek: oh okay. so i'll check to see if this has a big impact on our common API operations | 16:26 |
| mdbooth | Ok, looking | 16:26 |
| dims | mdbooth http://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2015-11-10.log.html#t2015-11-10T14:59:27 | 16:27 |
| *** jamielennox has quit IRC | 16:28 | |
| dims | mdbooth i see nova.tests.functional.test_server_group.ServerGroupTest.test_boot_servers_with_affinity and nova.tests.unit.api.ec2.test_cinder_cloud.CinderCloudTestCase.test_stop_with_attached_volume at the very least in the irc chat | 16:28 |
| mdbooth | dims: So, my patch should be producing log messages if it's causing long waits | 16:31 |
| mdbooth | dims: I don't know what I'm looking at in those irc logs | 16:35 |
| mdbooth | dims: So, hold on a sec | 16:36 |
| mdbooth | I'd like to make an alternate proposal | 16:36 |
| mdbooth | dims: There's a reasonable chance here that the underlying issue is actually a race in the tests | 16:37 |
| mdbooth | Which results in an ordering issue | 16:37 |
| mdbooth | The hard hang is something I experienced myself, which is why I wrote the tests to use eventlet explicitly | 16:38 |
| mdbooth | Because it doesn't hard hang | 16:38 |
| mdbooth | Of course, my tests don't hang at all | 16:38 |
| mdbooth | But when I was debugging, an eventlet hang responds to ctrl-c | 16:38 |
| mdbooth | In this case, the bug is most likely to be in the test, not in oslo.messaging | 16:39 |
| mdbooth | And the best way to find that would be to have it produce an error rather than hang | 16:39 |
| *** ansiwen has joined #openstack-oslo | 16:39 | |
| mdbooth | Which is what my follow-on patch does | 16:39 |
| mdbooth | https://review.openstack.org/#/c/238985/ | 16:39 |
| mdbooth | So, if we merged that with, say, a default timeout of 5 minutes | 16:40 |
| mdbooth | After we flush out all the problematic tests we can remove it | 16:40 |
| mdbooth | sdague: ^^^ ? | 16:40 |
| sdague | mdbooth: this is basically blocking all of nova dev, so it should be reverted and a reverted version released | 16:40 |
| sdague | then you can experiment on versions of this which fix it in a different way | 16:41 |
| mdbooth | sdague: I hear you. Is there another good way to find the buggy tests? | 16:41 |
| sdague | mdbooth: just run the nova unit tests locally | 16:41 |
| *** jamielennox has joined #openstack-oslo | 16:41 | |
| mdbooth | sdague: What proportion were failing? | 16:41 |
| sdague | I'm not convinced there is anything buggy about the tests | 16:42 |
| sdague | they are using the high level semantics that already exist | 16:42 |
| *** harlowja_at_home has quit IRC | 16:42 | |
| sdague | mdbooth: it's locally reproducable, if you would just do that, it would go a long way | 16:42 |
| *** salv-orlando has joined #openstack-oslo | 16:42 | |
| mdbooth | I literally heard about this about 5 minutes ago :) | 16:43 |
| mdbooth | sdague: Is it 100% reproducible locally? | 16:43 |
| mdbooth | Or is it non-deterministic? | 16:43 |
| sdague | we had 4 people working in channel that could all 100% locally reproduce this, and locally produced that the revert of your patch fixed it | 16:43 |
| sdague | it's 100% | 16:43 |
| mdbooth | Ok, cool | 16:43 |
| sdague | which is why it needs a revert | 16:43 |
| mdbooth | Well, that should be simple enough to sort out | 16:43 |
| mdbooth | Yup, I agree with a revert in that case | 16:44 |
| *** cdent has quit IRC | 16:44 | |
| mdbooth | I'll run nova tests locally with patched oslo.messaging | 16:44 |
| ansiwen | mdbooth: unit tests? | 16:46 |
| mdbooth | ansiwen: ? | 16:46 |
| mdbooth | Which unit tests? | 16:46 |
| ansiwen | mdbooth: you are going to run the unit tests locally? | 16:46 |
| ansiwen | because I just did that, and they were all fine | 16:47 |
| mdbooth | ansiwen: Hmm | 16:47 |
| mdbooth | So it's non-deterministic... | 16:47 |
| mdbooth | That's unfortunate | 16:47 |
| mdbooth | In that case, I'm not necessarily in favour of the revert | 16:47 |
| mdbooth | So, if you have a look at the original commit, even several of the oslo.messaging tests were calling server methods in the wrong order | 16:48 |
| ansiwen | I just did "tox -e py27" | 16:48 |
| mdbooth | ansiwen: Are you sure it was using the affected version of oslo.messaging? | 16:48 |
| ansiwen | mdbooth: no :-) | 16:49 |
| ansiwen | probably not... I already suspected that I was missing an important detail ;-) | 16:49 |
| *** itisha has joined #openstack-oslo | 16:52 | |
| sdague | ansiwen: .tox/py27/bin/pip freeze | grep oslo.messagin | 16:52 |
| mdbooth | sdague: Which is the bad version? | 16:54 |
| sdague | 2.8 | 16:55 |
| mdbooth | sdague: Oh, interesting. That's what I already have. | 16:55 |
| sdague | oslo.messaging==2.7.1.dev19 is the largest version that will pass | 16:55 |
| mdbooth | I'm convinced I've done several local runs. | 16:55 |
| sdague | mdbooth: and what's the max test time | 16:55 |
| mdbooth | I didn't pay any attention, tbh | 16:56 |
| mdbooth | But I will this time | 16:57 |
| sdague | mdbooth: right, that's the issue | 16:57 |
| mdbooth | sdague: What's a bad value? | 16:57 |
| sdague | a whole bunch of tests are going to 160s | 16:57 |
| sdague | the max test run time should be no more than 20s | 16:57 |
| sdague | that all stacks up and we overrun the overall test timeout | 16:57 |
| sdague | and explode | 16:57 |
| mdbooth | sdague: Got it. Do you have to hand a specific example of an overrunning test? | 16:57 |
| mdbooth | I can presumably reproduce by running all of them, but if you have one to hand... | 16:58 |
| sdague | mdbooth: here is a failed test run - http://logs.openstack.org/62/237762/8/check/gate-nova-python27/046e095/console.html there are items in there that hit that | 16:59 |
| mdbooth | Ah, I think dims posted a couple | 16:59 |
| sdague | when I said 100% repeatable, it's not that each test is 100% fail, it's that the effect is 100% repeatable on the nova unit tests as whole | 16:59 |
| mdbooth | sdague: Hmm, ok | 16:59 |
| sdague | run all the tests | 17:00 |
| sdague | tox -e py27 | 17:00 |
| *** bauzas has joined #openstack-oslo | 17:01 | |
| mdbooth | sdague: Have reproduced locally. | 17:02 |
| sdague | mdbooth: awesome, cool | 17:03 |
| mdbooth | sdague: I'm just pumping you for info, btw :) Assuming the revert will proceed full speed. | 17:03 |
| sdague | ok, no problem, I just got worried with all the language of being -1 on the revert | 17:04 |
| mdbooth | sdague: That was before I realised it was 100% reproducible. | 17:04 |
| mdbooth | I'm still expecting to find a problem elsewhere, I just didn't want to lose the opportunity. | 17:05 |
| mdbooth | If I can reproduce, that's not an issue. | 17:05 |
| sdague | yeh, given that we had a bunch of people reproducing it locally (myself included) it seemed like it would be easy enough to test future things | 17:06 |
| *** salv-orlando has quit IRC | 17:08 | |
| dims | mdbooth the +2A for revert went through before you joined the channel :) | 17:10 |
| mdbooth | dims: Glad to know where I stand :) | 17:11 |
| mdbooth | Yeah, totally makes sense now I have the context. | 17:11 |
| mdbooth | Ok, I have a suspect | 17:16 |
| *** pm90__ has quit IRC | 17:16 | |
| mdbooth | It's the cleanup of ServiceFixture | 17:16 |
| mdbooth | sdague dims: Fixed it | 17:20 |
| mdbooth | It's a bug in ServerGroupTest.setUp() | 17:20 |
| mdbooth | It calls self.start_service(), which adds a ServiceFixture | 17:20 |
| mdbooth | Which adds a cleanup action of kill, which calls stop | 17:21 |
| mdbooth | But it adds a second cleanup action | 17:21 |
| mdbooth | The second one hangs | 17:21 |
| mdbooth | Although, on second thoughts... should it? | 17:21 |
| * mdbooth ponders that for a minute | 17:21 | |
| sdague | so, iirc, that was the operation order that was required to not hang in the past | 17:23 |
| mdbooth | Hehe, fun | 17:23 |
| sdague | it's been a while since that was out there though | 17:23 |
| sdague | I think I built those fixtures a year ago | 17:23 |
| mdbooth | I'm just trying to work out why the second stop should hang | 17:24 |
| mdbooth | Because it seems it should be a no-op | 17:24 |
| mdbooth | And not hang | 17:24 |
| mdbooth | Unless something has subsequently called wait... | 17:24 |
| sdague | right, so I'm trying to remember what was happening the last time | 17:24 |
| sdague | I feel like what was going on was that there was an implied wait in the in memory driver on stop | 17:25 |
| *** yassine has quit IRC | 17:25 | |
| sdague | because it assumed there were still events to process | 17:25 |
| sdague | and if the queue was already empty then it would just hang | 17:25 |
| mdbooth | Ah... yes | 17:25 |
| mdbooth | It does: | 17:25 |
| mdbooth | self.rpcserver.stop() | 17:25 |
| mdbooth | self.rpcserver.wait() | 17:25 |
| mdbooth | dims harlowja: So this is the consequence of the wrapping decision ^^^ | 17:26 |
| mdbooth | sdague: The reason it hung is because it has stopped and waited, and now assumes it must be started again before you can call stop | 17:27 |
| *** dims has quit IRC | 17:27 | |
| mdbooth | sdague: So, that genuinely is a semantic change in oslo.messaging, but it was a deliberate one | 17:27 |
| mdbooth | Which now looks misguided :/ | 17:27 |
| *** dims has joined #openstack-oslo | 17:28 | |
| *** browne has joined #openstack-oslo | 17:28 | |
| mdbooth | dims: Did you see above before you quit, btw? | 17:28 |
| sdague | ok, yeh, that kind of semantic change is going to need a lot of warning, and might have more unintended downstream effects than this | 17:28 |
| mdbooth | sdague: The second kill should also go, I think, but it feels wrong that we didn't handle it | 17:29 |
| *** achanda has joined #openstack-oslo | 17:30 | |
| * mdbooth will send a mail to the list | 17:31 | |
| mdbooth | sdague: Sorry for the trouble | 17:31 |
| sdague | mdbooth: yeh, no problem. Sorry for any frustration. This ended up being the 3rd conversation about the same bug today :) | 17:31 |
| openstackgerrit | Akihiro Motoki proposed openstack/oslo.config: Add max length check to StrOpt https://review.openstack.org/242880 | 17:32 |
| dims | sdague mdbooth i think i caught most of it. apologies for the network reconnects | 17:34 |
| mdbooth | dims: In case you missed the critical bit, the issue is that a set of nova tests is doing: | 17:34 |
| mdbooth | server.start(), server.stop(), server.wait(), server.stop(), server.wait() | 17:35 |
| mdbooth | Because we decided to allow a server to be restarted, after wait() we're now expecting another start() | 17:35 |
| dims | mdbooth do we log anything in the step #4? | 17:35 |
| mdbooth | dims: Good question, and I'll go look | 17:35 |
| dims | "can't stop something that has not been started" | 17:36 |
| mdbooth | Yup | 17:36 |
| *** achanda has quit IRC | 17:36 | |
| mdbooth | So the second call to stop() hangs because it's expecting a start | 17:36 |
| mdbooth | If we *don't* allow a server to be restarted, this goes away | 17:36 |
| mdbooth | Both the stop() and the wait() become no-ops | 17:36 |
| dims | but then we have to fix folks who are reusing the same thing | 17:36 |
| mdbooth | True. Is anybody doing that? | 17:37 |
| dims | mdbooth if we make a release we'll find out...but don't want to take that chance :) | 17:37 |
| mdbooth | Either way, I'm not sure it's possible to completely fix both things :) | 17:37 |
| dims | let's get sileht's opinion too. | 17:38 |
| mdbooth | dims: How many projects use oslo.messaging? | 17:39 |
| mdbooth | We could just run all their tests? | 17:39 |
| mdbooth | dims: I'll send an email | 17:41 |
| mdbooth | Probably easier to do on the list, and no great emergency, I think | 17:41 |
| dims | mdbooth http://paste.openstack.org/show/478471/ | 17:41 |
| dims | this is just from my local directories | 17:41 |
| mdbooth | That's doesn't look intractable | 17:42 |
| mdbooth | Is there a way we can get jenkins to run them? | 17:42 |
| mdbooth | Submit a patch to requirements.txt, I guess | 17:42 |
| mdbooth | That should run all their tests, no? | 17:43 |
| dims | wish it was that simple | 17:43 |
| mdbooth | However, that would also require a release... | 17:43 |
| mdbooth | Is there a way to get them to run against an arbitrary commit? | 17:43 |
| mdbooth | I'm sure I saw you submit some arcane invocation to nova once. | 17:44 |
| dims | mdbooth y, if it was one project... | 17:45 |
| dims | that's why i had the travis setup...looks like the new install_command(s) broke what i had | 17:46 |
| dims | (for unit tests) | 17:46 |
| mdbooth | The problem is, the previous behaviour was basically undefined | 17:52 |
| mdbooth | dims: Incidentally, yes the warning message are emitted | 17:52 |
| mdbooth | Somewhere | 17:52 |
| *** pm90_ has joined #openstack-oslo | 17:55 | |
| * mdbooth is interested to know why the test ever completes | 17:57 | |
| *** SurajD has joined #openstack-oslo | 17:59 | |
| *** jecarey has quit IRC | 17:59 | |
| *** browne has quit IRC | 18:01 | |
| harlowja | mdbooth soo its stop, wait, start ... (and repeat) stuff showing up? | 18:03 |
| * harlowja just cathing up | 18:04 | |
| mdbooth | harlowja: It's start, stop, wait, stop, wait | 18:04 |
| harlowja | ah | 18:04 |
| mdbooth | I'm just writing it up in a mail | 18:04 |
| harlowja | kk | 18:04 |
| harlowja | stop wait stop stop stop wait wait wait | 18:04 |
| harlowja | just messing with u (continue writing the email) | 18:04 |
| harlowja | lol | 18:04 |
| *** jerrygb has joined #openstack-oslo | 18:06 | |
| mdbooth | sdague: Incidentally, any idea where the logs for those test runs are going? It should have emitted a log message after 30 seconds, and in my tracing it does exactly that. However, I didn't see the logs anywhere. The log message is both a message at WARN, and a full stack trace at DEBUG. | 18:09 |
| *** jerrygb_ has quit IRC | 18:09 | |
| *** sputnik13 has joined #openstack-oslo | 18:10 | |
| *** sputnik13 has quit IRC | 18:11 | |
| *** sputnik13 has joined #openstack-oslo | 18:11 | |
| sdague | they might be in the subunit stream | 18:11 |
| sdague | debug is definitely off | 18:11 |
| sdague | because it creates too much output | 18:11 |
| sdague | http://logs.openstack.org/62/237762/8/check/gate-nova-python27/046e095/tmpfFPVIr is the raw subunit stream for the failed test run I posted above | 18:13 |
| *** pm90__ has joined #openstack-oslo | 18:13 | |
| *** SurajD has quit IRC | 18:14 | |
| *** SurajD has joined #openstack-oslo | 18:14 | |
| SurajD | Can somebody review this https://review.openstack.org/#/c/240661/ this is my first upstream contribution | 18:16 |
| *** pm90_ has quit IRC | 18:16 | |
| dims | SurajD one nit left by gcb...i'll +2 after that | 18:17 |
| dims | SurajD welcome! :) | 18:17 |
| SurajD | dims, sure | 18:18 |
| SurajD | dims, thanks :) | 18:18 |
| mdbooth | sdague: Yup, they're in there. Thanks. | 18:18 |
| mdbooth | There are 8 | 18:19 |
| openstackgerrit | Suraj Deshmukh proposed openstack/oslo.utils: Added ICMP 'type' and 'code' checking capability to 'netutils' module https://review.openstack.org/240661 | 18:20 |
| *** abitha has joined #openstack-oslo | 18:22 | |
| SurajD | dims, done with nit :) | 18:22 |
| dims | cool, Happy Diwali SurajD | 18:23 |
| SurajD | dims, Thank you and same to you :) | 18:23 |
| *** achanda has joined #openstack-oslo | 18:25 | |
| SurajD | dims, Thanks for the Code-review as a DIwali Gift | 18:28 |
| harlowja | lol | 18:28 |
| *** bnemec has quit IRC | 18:29 | |
| mdbooth | harlowja: Thar she blows | 18:31 |
| harlowja | kk, let's seeee here | 18:32 |
| *** rlrossit has left #openstack-oslo | 18:32 | |
| mdbooth | harlowja: I've blown so far past going home time my wife just turned up with my plated dinner covered in tinfoil :) | 18:33 |
| mdbooth | I'm happy to deal with this async. Catch you tomorrow, and thanks for looking. | 18:33 |
| openstackgerrit | Merged openstack/oslo.messaging: Revert "Robustify locking in MessageHandlingServer" https://review.openstack.org/243658 | 18:34 |
| harlowja | mdbooth np | 18:36 |
| dims | mdbooth talk to you tomorrow | 18:37 |
| dims | harlowja does this look ok to you? https://review.openstack.org/#/c/243275/ | 18:37 |
| harlowja | dims looks ok to me, checked hash and thats the one with the change | 18:39 |
| dims | harlowja mind +1'ing it please? | 18:39 |
| harlowja | kk | 18:39 |
| *** achanda has quit IRC | 18:41 | |
| rjaiswal | thanks harlowja, dims | 18:42 |
| harlowja | np | 18:42 |
| *** yamahata has joined #openstack-oslo | 18:45 | |
| *** SurajD has quit IRC | 18:51 | |
| *** mriedem has quit IRC | 18:53 | |
| *** jecarey_ has joined #openstack-oslo | 18:57 | |
| *** jecarey_ has quit IRC | 18:57 | |
| *** jecarey_ has joined #openstack-oslo | 18:57 | |
| *** mriedem has joined #openstack-oslo | 18:58 | |
| sputnik13 | dims have you tried bringing up devstack with zookeeper lately? | 18:59 |
| *** browne has joined #openstack-oslo | 18:59 | |
| dims | sputnik13 not for a few days, what's up? | 18:59 |
| sputnik13 | we've been using zookeeper in devstack and it seems like our jobs fail since the zookeeper addition to devstack | 19:00 |
| sputnik13 | we had our own devstack plugin for zookeeper | 19:00 |
| sputnik13 | so I removed our plugin and the gate still fails, just wondering whether it's just me | 19:00 |
| dims | got a review where this is failing? | 19:01 |
| *** jerrygb_ has joined #openstack-oslo | 19:01 | |
| harlowja | sputnik13 dims a coworker who runs stuff on rhel7 is also investigating it there and figuring out whats packaged and whats not, asked him to send an email to openstack-dev with his results | 19:02 |
| sputnik13 | http://logs.openstack.org/66/243366/1/check/gate-cue-integration-dsvm-rabbitmq/5f382f5/logs/devstacklog.txt.gz#_2015-11-10_01_07_45_606 | 19:02 |
| *** achanda has joined #openstack-oslo | 19:03 | |
| *** jerrygb has quit IRC | 19:04 | |
| harlowja | hmmm | 19:05 |
| harlowja | my guess is thats because the infra servers are already starting zookeeper | 19:05 |
| *** jerrygb has joined #openstack-oslo | 19:05 | |
| harlowja | and devstack should have some smarts to not bother starting zookeeper if its already there and running | 19:05 |
| harlowja | *infra servers / infra vms | 19:06 |
| sputnik13 | meh? | 19:06 |
| dims | harlowja i thought there was a review already for that | 19:07 |
| harlowja | unsure | 19:07 |
| harlowja | sputnik13 https://review.openstack.org/#/c/241040/ (See second comment) | 19:07 |
| harlowja | but dims might know of an existing patch already in the pipeline | 19:07 |
| harlowja | arun (also commented there, a coworker) is emailing the rhel7/fedora people to figure out about rhel7 zookeeper package (which appears underway) | 19:08 |
| *** jerrygb_ has quit IRC | 19:08 | |
| dims | sputnik13 https://review.openstack.org/#/c/242445/ | 19:08 |
| *** achanda has quit IRC | 19:09 | |
| sputnik13 | yeah we circumvented this whole deal by just not starting zookeeper | 19:09 |
| sputnik13 | since it gets started automatically when you install the package | 19:10 |
| harlowja | right | 19:10 |
| sputnik13 | https://github.com/openstack/cue/blob/master/contrib/devstack/lib/zookeeper#L47 | 19:11 |
| harlowja | cool, so ^ might not be needed anymore | 19:11 |
| harlowja | ' sudo pip install -e "git+https://github.com/python-zk/kazoo.git#egg=kazoo"' | 19:11 |
| harlowja | :-/ | 19:11 |
| harlowja | hmmm | 19:12 |
| harlowja | what is this, ha | 19:12 |
| harlowja | https://github.com/openstack/cue/blob/master/contrib/devstack/lib/zookeeper#L40 ;) | 19:12 |
| *** achanda has joined #openstack-oslo | 19:12 | |
| sputnik13 | well, yeah so that cue review is to remove that, but the gate still fails because the current zookeeper handler in devstack tries to start zookeeper | 19:12 |
| harlowja | guess that was to get the slow-butt-kaozo-releases | 19:12 |
| sputnik13 | https://review.openstack.org/#/c/243366/ | 19:12 |
| harlowja | cool | 19:12 |
| dims | sputnik13 you can try Depends-On the review that i pointed to | 19:13 |
| *** achanda has quit IRC | 19:13 | |
| sputnik13 | dims the thing is none of our checkins will go through until this is resolved since zookeeper is a required component | 19:13 |
| sputnik13 | what's the likelihood this will be merged some time soon | 19:14 |
| sputnik13 | does systemd return "running"? | 19:15 |
| sputnik13 | I forget | 19:15 |
| sputnik13 | err when you do a service status | 19:15 |
| harlowja | $ service zookeeper status | 19:18 |
| harlowja | zookeeper start/running, process 1176 | 19:18 |
| harlowja | so seems so | 19:18 |
| *** sputnik13 has quit IRC | 19:18 | |
| harlowja | $ service zookeeper status | 19:18 |
| harlowja | zookeeper stop/waiting | 19:18 |
| harlowja | 19:18 | |
| harlowja | ^ for when stopped | 19:18 |
| harlowja | thats on Ubuntu 14.04.1 LTS | 19:18 |
| dims | dhellmann, lifeless : nova needs to get started on using oslo.versionedobjects[fixtures], can you please review https://review.openstack.org/#/c/238871/ once more when you get a chance | 19:19 |
| dims | mriedem ^^ | 19:19 |
| *** sputnik13 has joined #openstack-oslo | 19:19 | |
| dims | mriedem know anyone else who has project-config karma? | 19:19 |
| mriedem | AJaeger | 19:20 |
| mriedem | i'm trying out a hack in tox.ini for now | 19:20 |
| mriedem | just adding oslo.versionedobjects[fixtures] to deps | 19:21 |
| mriedem | so it's not in test-requirements.txt | 19:21 |
| *** eezhova has quit IRC | 19:21 | |
| dims | mriedem right. | 19:21 |
| sdague | honestly, I think in the zookeeper case the contract is just wrong | 19:21 |
| sdague | with all other services we do restart as our contract | 19:22 |
| openstackgerrit | Joshua Harlow proposed openstack/oslo.utils: Add useful 'time_it' decorator https://review.openstack.org/231220 | 19:22 |
| *** eezhova has joined #openstack-oslo | 19:23 | |
| dims | sdague so start_service -> restart_service? | 19:23 |
| sdague | yeh, so, actually, let me take a whack at abstracting this a little | 19:24 |
| harlowja | btw, https://admin.fedoraproject.org/pkgdb/package/zookeeper/timeline#n1 (also is happening) | 19:24 |
| harlowja | 'user: ctubbsii requested branch: epel7 for package zookeeper' ... | 19:24 |
| dims | sdague ack thanks, some services we using restart_service, some start_service, i just happened to pick one :) | 19:24 |
| sdague | yeh | 19:24 |
| sputnik13 | why should start_service == restart_service? | 19:25 |
| *** ihrachys has quit IRC | 19:25 | |
| dims | sputnik13 apparently zookeeperd is already in the CI image deployed, so zookeeper starts even before devstack is installed | 19:25 |
| *** yamahata has quit IRC | 19:25 | |
| *** yamahata has joined #openstack-oslo | 19:26 | |
| sputnik13 | I agree it's inconsistent with what the rest of devstack may be doing, but start_service == restart_service intuitively is wrong to me | 19:26 |
| dims | sputnik13 that's why i did not do it :) | 19:26 |
| *** bauzas has left #openstack-oslo | 19:36 | |
| openstackgerrit | Joshua Harlow proposed openstack/taskflow: Move engine options extraction to __init__ methods https://review.openstack.org/243793 | 19:38 |
| sputnik13 | so uhh... how does one go about getting their project or gate job added to the devstack project's gate list so that changes like these don't just pull out the rug from under us? | 19:38 |
| sputnik13 | is that even an option? | 19:38 |
| *** subscope has joined #openstack-oslo | 19:39 | |
| harlowja | unsure sdague might know? | 19:39 |
| harlowja | (or others on infra channel) | 19:40 |
| *** amotoki has quit IRC | 19:42 | |
| sdague | sputnik13: no, we're not gating the world here. It's about engaging once issues show up | 19:42 |
| sdague | dims / harlowja - how do you feel about this approach - https://review.openstack.org/243794 | 19:43 |
| sputnik13 | gate check failed http://logs.openstack.org/45/242445/2/check/gate-tempest-dsvm-neutron-full/c8f0736/console.html#_2015-11-10_19_26_25_554 | 19:43 |
| sdague | sputnik13: yep, that's probably the 35% failure rate issue I posted on the ML | 19:43 |
| harlowja | sdague ok, thats fine with me, we might want to push on the epel7 zookeeper change though (or maybe not?) | 19:44 |
| sdague | harlowja: what change is this? | 19:44 |
| sputnik13 | sdague that assumes dlm is the only user for zookeeper, dlm is a library that uses zookeeper | 19:44 |
| harlowja | sdague https://admin.fedoraproject.org/pkgdb/package/zookeeper/timeline#n1 | 19:44 |
| harlowja | epel7 process for packaging zookeeper for rhel7 started... | 19:44 |
| sputnik13 | for services that use zookeeper doing "install_dlm" when we want zookeeper seems weird | 19:45 |
| harlowja | (its in fedora already, just someone needs to move it to epel7...) | 19:45 |
| harlowja | sdague that was 4 days ago (not sure how long it takes) | 19:45 |
| sdague | sputnik13: so, the point of the exercise on using tooz is that we should make sure all the zookeeper interaction, for any purpose, goes through tooz | 19:46 |
| sdague | which probably means extending tooz to support whatever that set of semantics is | 19:46 |
| sputnik13 | sdague once again, you're assume zookeeper's only use is for dlm, this is not true | 19:46 |
| sdague | sputnik13: no, I really don't | 19:46 |
| sputnik13 | taskflow uses zookeeper for jobboard | 19:46 |
| * harlowja will have to figure out what tooz apis are missing for the jobboard stuff | 19:47 | |
| sputnik13 | monasca uses zookeeper for kafka | 19:47 |
| harlowja | i was thinking about that in the shower this morning | 19:47 |
| harlowja | and nuff said | 19:47 |
| harlowja | lol | 19:47 |
| dims | too much information harlowja | 19:47 |
| sputnik13 | yeah I don't really want to know what you do in the shower | 19:47 |
| sputnik13 | :) | 19:47 |
| harlowja | lol | 19:47 |
| sdague | sputnik13: right but if we end up with an environment that's consul & zookeeper to be functional, than this whole thing failed | 19:48 |
| dims | sdague i am ok with the lib/dlm change | 19:48 |
| harlowja | http://docs.openstack.org/developer/taskflow/jobs.html (for those who wondering what sputnik13 is talking about) | 19:48 |
| sdague | dims: cool, just had to fix a set of includes there | 19:48 |
| harlowja | i'll eventually i guess move the drivers there to tooz (sometime in the future) | 19:49 |
| harlowja | i think that makes logical sense to try to do (if we can) | 19:49 |
| dims | sdague some comments at the top with the method names needs fixing | 19:49 |
| harlowja | http://docs.openstack.org/developer/taskflow/jobs.html#taskflow.jobs.backends.impl_zookeeper.ZookeeperJobBoard also has other comments about how it works... | 19:49 |
| sputnik13 | so are we now saying that anything that uses zookeeper directly can't be used in openstack? | 19:49 |
| harlowja | lol | 19:50 |
| harlowja | i hope not | 19:50 |
| harlowja | cause thats cray cray imho | 19:50 |
| sputnik13 | I don't get the need to say zookeeper falls strictly under tooz/dlm | 19:50 |
| sdague | ok, I was not in whatever room the conversations happened in | 19:50 |
| sdague | but there seems to be a missing piece of all of this conversation | 19:51 |
| sdague | because this was what projects could assume would always be there, and the abstraction of tooz was picked | 19:52 |
| sdague | zookeeper direct was not picked | 19:52 |
| sputnik13 | sdague I believe the abstraction of tooz as DLM was the decision | 19:52 |
| sdague | if there are other voices that want zookeeper direct picked, they should voice that fact | 19:52 |
| sputnik13 | not that tooz is an abstraction for all zookeeper things | 19:52 |
| sdague | sputnik13: and that a dlm could be assumed | 19:52 |
| harlowja | projects already exist and those things use zookeeper, so we likely need to accept that it will be there in those projects (especially ones we don't control like kafka and others) and for ones we can affect, they should use tooz (if the api works, or can be adjusted to work) | 19:52 |
| harlowja | thats my 2 cents | 19:53 |
| sdague | there is no assumption that you can assume zookeeper is there, or that people will deploy zookeeper | 19:53 |
| sputnik13 | so anyone needing a DLM like thing must use tooz, that's accepted and agreed, I'm one of the proponents of the idea | 19:53 |
| sdague | so any use of that in a project probably needs to be optional | 19:53 |
| harlowja | sure, if u don't want to use a feature of a project, well sucks for that person, lol | 19:53 |
| harlowja | *and/or optional feature | 19:54 |
| sputnik13 | I'm not questioning whether tooz should be the standard interface for DLM | 19:54 |
| openstackgerrit | Ryan McNair proposed openstack/oslo.concurrency: Allow deletion of all lock files matching a glob https://review.openstack.org/241663 | 19:54 |
| sputnik13 | what I'm saying is, there are existing projects that use zookeeper for something other than DLM, the addition of this "dlm support" to devstack broke said project, which is all fine we're for sharing code and removing duplication so we want to remove our own zookeeper devstack/lib thing and use the common one | 19:56 |
| *** rlrossit has joined #openstack-oslo | 19:56 | |
| sputnik13 | but within devstack putting zookeeper strictly under "dlm" seems weird to me since that's just a service that's used by dlm/tooz and now we have assertions being made that zookeeper should never be used directly by anything which is completely beyond the scope of what the DLM conversation/discussion was bout | 19:58 |
| dims | i see the point sputnik13 | 19:59 |
| dims | not sure how we went from start/restart to this discussion :) | 20:00 |
| sputnik13 | yeah, seems a really windy path doesn't it? :) | 20:00 |
| openstackgerrit | Michael Krotscheck proposed openstack/oslo.middleware: Enable latent CORS configuration via pastedeploy https://review.openstack.org/240984 | 20:03 |
| *** yamahata has quit IRC | 20:11 | |
| *** tristanC has joined #openstack-oslo | 20:13 | |
| tristanC | Greeting oslo folks, I'd like to check in on bug 1449062. e.g., can oslo provides a command line utility "prlimit" to accomodate system that does not include util-linux v2.21 (which provide such utility) ? | 20:15 |
| openstack | bug 1449062 in Glance "qemu-img calls need to be restricted by ulimit (CVE-2015-5162)" [High,In progress] https://launchpad.net/bugs/1449062 - Assigned to nikhil komawar (nikhil-komawar) | 20:15 |
| *** lifeless has quit IRC | 20:26 | |
| *** rjaiswal has quit IRC | 20:26 | |
| *** rjaiswal has joined #openstack-oslo | 20:27 | |
| *** sputnik13 has quit IRC | 20:38 | |
| openstackgerrit | Joshua Harlow proposed openstack/taskflow: Move engine options extraction to __init__ methods https://review.openstack.org/243793 | 20:40 |
| openstackgerrit | Joshua Harlow proposed openstack/taskflow: Move engine options extraction to __init__ methods https://review.openstack.org/243793 | 20:41 |
| *** pm90__ has quit IRC | 20:42 | |
| *** sputnik13 has joined #openstack-oslo | 20:42 | |
| haypo | tristanC: where do you want to put such helper script? | 20:44 |
| tristanC | haypo: preferably in $PATH | 20:46 |
| tristanC | haypo: it will be used by glance, cinder and nova, right before each call to qemu-img info actually | 20:46 |
| haypo | tristanC: no, i'm asking in which oslo library. there is no such "oslo" library anymore :) | 20:46 |
| haypo | oslo.concurrency maybe, there is already a processutils.py | 20:47 |
| tristanC | haypo: yes, that would be the most logical place | 20:47 |
| *** bnemec has joined #openstack-oslo | 20:47 | |
| haypo | tristanC: is there a patch somewhere for oslo.concurrency? | 20:48 |
| tristanC | haypo: no, not yet. I'd have to check how a new script can be installed by a library first | 20:49 |
| haypo | tristanC: i never used the prlimit command limit tool. it looks like it changes the limit after the process was started | 20:49 |
| tristanC | haypo: that's it, the code will be pretty straightforward, getopt into resource.setrlimit calls before doing the execve | 20:50 |
| haypo | oh, right. currently, setup.cfg of oslo.utils doesn't have any entry point | 20:50 |
| haypo | do you really need a command line script if it's written in python? maybe you can use python -c 'import ...; main()' arg1 arg2 ... | 20:51 |
| *** sputnik13 has quit IRC | 20:51 | |
| *** jerrygb has quit IRC | 20:51 | |
| tristanC | hum, the proposed idea is to provide a fallback if prlimit is not already installed | 20:51 |
| *** sputnik13 has joined #openstack-oslo | 20:52 | |
| tristanC | (and prlimit is not installed in most LTS system because it needs util-linux-v2.21 | 20:53 |
| haypo | tristanC: why not calling setrlimit(RLIMIT_AS) and then exec()? | 20:53 |
| haypo | tristanC: prlimit call may occur too late, after the child process already killed the memory, no? | 20:53 |
| tristanC | haypo: ha, that's what we tried, but then you call the setrlimit on a process that may already have exceed the limit, resulting in a harakiri type of outcomes... | 20:54 |
| tristanC | haypo: while a command line utility will apply resource limitation on a controlled environment (post execve) | 20:55 |
| haypo | tristanC: i'm not proposing to do that after a fork, but after a fork+exec, so basically in a new program | 20:55 |
| harlowja | isn't there a post execv hook already in python? | 20:55 |
| haypo | tristanC: i'm asking why you really want the prlimit program, it doesn't seem safe | 20:55 |
| tristanC | harlowja: not that I'm aware of | 20:56 |
| haypo | harlowja: nope. subprocess provides preexec_fn which occurs between fork and exec, it's different | 20:56 |
| harlowja | haypo kk, ya, just checked, preexec, not post | 20:56 |
| harlowja | but why not use preexec to set limits? | 20:56 |
| harlowja | 'this object will be called in the child process just before the child is executed. ' | 20:57 |
| harlowja | (just a thought) | 20:57 |
| tristanC | haypo: if we call "setrlimit(RLIMIT_AS) and then exec()", then you might not get to the exec if the process already exceed memory limitation | 20:57 |
| tristanC | haypo: fork is not enough since both process will inherit heap/stack/... We really needs a fork->execve(prlimit)->execve(target) | 20:58 |
| harlowja | qemu-img need a python api, lol | 20:58 |
| haypo | tristanC: which would a fresh process uses more than ... say 50 MB? | 20:58 |
| haypo | tristanC: i'm not sure that you understood what i proposed | 20:58 |
| haypo | tristanC: i really propose to write an helper program | 20:58 |
| *** jerrygb has joined #openstack-oslo | 20:59 | |
| *** jerrygb has quit IRC | 20:59 | |
| haypo | tristanC: nova spawns the command "wrapper --limit=1G qemu-img", wrapper sets RLIMIT_AS and call exec(qemu...) | 20:59 |
| harlowja | isn't that what preexec_fn can do? | 20:59 |
| haypo | spawns: subprocess.call(...), so run a new process using fork+exec | 20:59 |
| tristanC | haypo: oh right! then what you describe is the "prlimit" utility | 20:59 |
| *** jerrygb has joined #openstack-oslo | 20:59 | |
| haypo | harlowja: nope. see what tristanC explained before, it doesn't work. for example, if nova uses 10G, you cannot set the limit to 1G | 20:59 |
| *** amrith is now known as _amrith_ | 21:00 | |
| harlowja | nova is using 10G :-/ | 21:00 |
| harlowja | i hope not, lol | 21:00 |
| haypo | tristanC: oh. i only saw the first usage of prlimit in the man page :) so yes, it can also run a command | 21:00 |
| haypo | harlowja: it was just an example. preexec_fn doesn't work, see the bug | 21:01 |
| harlowja | haypo ok ok | 21:01 |
| tristanC | harlowja: long story short, I proposed that, it got merged since it passed nova gate, but then fwaas gate made nova process baloon >2Go of ram and the preexec trick broke their gate, patch got reverted in less than 24hours :) | 21:01 |
| haypo | tristanC: so yes, i propose to reimplement a simple prlimit in pure python | 21:01 |
| haypo | tristanC: and put this in oslo.concurrency | 21:01 |
| tristanC | haypo: awesome, do you want me to propose such change ? | 21:01 |
| haypo | tristanC: in term of API, it can be a new parameter passed to execute(), at least for the end user of the API | 21:02 |
| haypo | tristanC: about the need of creating a script: we can use python -c '...code...', but i checked, python -m module ... also works on py2 | 21:02 |
| haypo | (but it's more limited) | 21:02 |
| haypo | tristanC: try: echo data|python -m base64 | 21:03 |
| tristanC | haypo: well first we need a prlimit in pure python, then either a prlimit python helper to decide whenever use the nativ implementation or the full python; | 21:03 |
| tristanC | haypo: and then, yes, a "resources_limit" parameter to the execute method would be neat | 21:03 |
| haypo | tristanC: hum, i would prefer to keep it simple. so support RLIMIT_AS | 21:03 |
| haypo | (only support RLIMIT_AS) | 21:04 |
| haypo | it can be extended later | 21:04 |
| tristanC | haypo: works for me, the python oneliner sounds like a neat trick | 21:05 |
| haypo | haypo@smithers$ python -m oslo_concurrency.prlimit | 21:05 |
| haypo | ('hello world', ['/home/haypo/prog/openstack/oslo.concurrency/oslo_concurrency/prlimit.py']) | 21:05 |
| haypo | ah yes, it works :) | 21:05 |
| haypo | tristanC: i would prefer to avoid entry points in oslo.concurrency setup.cfg. it's a little bit strange to install a script when you install a library, no? | 21:05 |
| tristanC | much better than worying about adding a new script to the toolchain | 21:05 |
| haypo | (at least, unusual) | 21:05 |
| tristanC | ok, I'll propose something along those lines for oslo.concurency | 21:06 |
| haypo | tristanC: so yes, i'm writing a PoC right now | 21:07 |
| *** lifeless has joined #openstack-oslo | 21:10 | |
| *** salv-orlando has joined #openstack-oslo | 21:10 | |
| *** yamamoto has joined #openstack-oslo | 21:11 | |
| *** jerrygb_ has joined #openstack-oslo | 21:12 | |
| *** jerrygb has quit IRC | 21:16 | |
| *** jerrygb_ is now known as jerrygb | 21:17 | |
| *** yamamoto has quit IRC | 21:18 | |
| tristanC | haypo: oh well, if you've something, please submit it, I'll propose Depends-On change for nova to get the full picture | 21:22 |
| tristanC | the former change was https://review.openstack.org/#/c/209627/ | 21:23 |
| *** jamielennox is now known as jamielennox|away | 21:25 | |
| openstackgerrit | Victor Stinner proposed openstack/oslo.concurrency: PoC: add memory_limit parameter to execute() https://review.openstack.org/243829 | 21:27 |
| haypo | tristanC: ^^ here is a short PoC written in 30 min | 21:28 |
| *** kgiusti has quit IRC | 21:28 | |
| tristanC | haypo: reading... ( https://review.openstack.org/#/c/243829/1/oslo_concurrency/prlimit.py ) | 21:29 |
| haypo | hum, maybe "memlimit.py" is a better name, since it only supports a memory limit | 21:29 |
| haypo | we may also make it private, to make it more explicit that it must be not used directly, but through execute() | 21:29 |
| tristanC | haypo: we also need RLIMIT_CPU | 21:30 |
| haypo | tristanC: ah? | 21:30 |
| haypo | tristanC: is it reliable? | 21:30 |
| haypo | tristanC: last time i implemented a sandbox (sic), i implemented the timeout in the parent process | 21:30 |
| tristanC | haypo: it is very reliable, set this to 2 and it will prevent more than 2 second of CPU time (not real time) | 21:31 |
| tristanC | haypo: which is good mitigation against infinit loop | 21:31 |
| *** itisha has quit IRC | 21:31 | |
| haypo | tristanC: you get a singal. can you catch the signal? | 21:31 |
| tristanC | haypo: it should make the process exit with a return code != 0, leading to execute() raising an exception | 21:33 |
| tristanC | haypo: the child process shouldn't be able to catch that signal | 21:33 |
| *** yamahata has joined #openstack-oslo | 21:33 | |
| haypo | oh, it's SIGKILL. i expected SIGXCPU | 21:35 |
| haypo | tristanC: well, i wrote the PoC to show what i had in mind. do you like such API and the design? (wrapper written in pure python) | 21:36 |
| tristanC | haypo: it's looking good to me. However nova dev would like to use the native prlimit tool if available | 21:38 |
| tristanC | which should be easy to implement in your design | 21:38 |
| *** sputnik13 has quit IRC | 21:42 | |
| haypo | tristanC: i'm not sure that a wrapper written in pure python is something secure, i mean you have to trust the PYTHONPATH environment variable and python modules installed in the system | 21:42 |
| haypo | tristanC: but i don't know neither if it's a concern for you :) | 21:43 |
| *** sputnik13 has joined #openstack-oslo | 21:43 | |
| haypo | python3 provides a nice -I option which "isolates" python to be a little bit more secure, but it may break setup where oslo.concurrency is not installed "as expected" | 21:43 |
| tristanC | haypo: that's fine, the issue here is about malicious inputs consumed by qemu-img, the host sanity and PYTHONPATH are out of scope | 21:45 |
| haypo | tristanC: right | 21:46 |
| *** thumpba has quit IRC | 21:54 | |
| *** yamahata has quit IRC | 21:56 | |
| *** thangp has quit IRC | 21:57 | |
| *** jerrygb has quit IRC | 21:57 | |
| haypo | modify a docstring and see integration tests failing: done :-) https://review.openstack.org/#/c/243630/ | 21:58 |
| openstackgerrit | Matt Riedemann proposed openstack/oslo.versionedobjects: Remove remote_object_calls from _BaseTestCase https://review.openstack.org/243840 | 21:58 |
| *** salv-orlando has quit IRC | 22:01 | |
| *** pradk has quit IRC | 22:03 | |
| *** yamamoto has joined #openstack-oslo | 22:07 | |
| *** salv-orlando has joined #openstack-oslo | 22:10 | |
| openstackgerrit | Merged openstack/oslo-incubator: tests: remove unused testmods https://review.openstack.org/243579 | 22:11 |
| openstackgerrit | John Eckersberg proposed openstack/oslo.messaging: rabbit: Add rabbit_queue_ttl option https://review.openstack.org/243845 | 22:12 |
| *** salv-orlando has quit IRC | 22:13 | |
| openstackgerrit | Joshua Harlow proposed openstack/taskflow: Move engine options extraction to __init__ methods https://review.openstack.org/243793 | 22:14 |
| *** _amrith_ is now known as amrith | 22:15 | |
| *** jamielennox|away is now known as jamielennox | 22:24 | |
| *** pm90_ has joined #openstack-oslo | 22:24 | |
| *** pm90__ has joined #openstack-oslo | 22:26 | |
| openstackgerrit | Matt Riedemann proposed openstack/oslo.versionedobjects: Move compare_obj to the fixture module for external consumption https://review.openstack.org/243852 | 22:26 |
| mriedem | dansmith: there we go ^ | 22:26 |
| dansmith | cool | 22:26 |
| openstackgerrit | Merged openstack/oslo.utils: Added ICMP 'type' and 'code' checking capability to 'netutils' module https://review.openstack.org/240661 | 22:28 |
| *** pm90_ has quit IRC | 22:28 | |
| *** boris-42 has joined #openstack-oslo | 22:30 | |
| *** nikhil_k has quit IRC | 22:33 | |
| *** regXboi has quit IRC | 22:33 | |
| *** nikhil has joined #openstack-oslo | 22:33 | |
| *** pm90_ has joined #openstack-oslo | 22:34 | |
| *** pm90__ has quit IRC | 22:37 | |
| *** pm90_ is now known as pratikmallya | 22:37 | |
| *** sputnik13 has quit IRC | 22:41 | |
| *** sputnik13 has joined #openstack-oslo | 22:41 | |
| *** bana_k has joined #openstack-oslo | 22:52 | |
| *** cprmrf__ has quit IRC | 22:54 | |
| openstackgerrit | Banashankar k proposed openstack/oslo.messaging: Fixing the server example code Added server.stop() before server.wait() https://review.openstack.org/243294 | 22:58 |
| *** rjaiswal has quit IRC | 23:05 | |
| openstackgerrit | Joshua Harlow proposed openstack/taskflow: Move engine options extraction to __init__ methods https://review.openstack.org/243793 | 23:06 |
| *** jerrygb has joined #openstack-oslo | 23:08 | |
| *** sputnik13 has quit IRC | 23:11 | |
| *** rlrossit has left #openstack-oslo | 23:14 | |
| *** dims has quit IRC | 23:15 | |
| *** jecarey_ has quit IRC | 23:15 | |
| *** sputnik13 has joined #openstack-oslo | 23:16 | |
| *** fultonj has joined #openstack-oslo | 23:20 | |
| *** fultonj has left #openstack-oslo | 23:20 | |
| *** pballand has quit IRC | 23:21 | |
| *** jecarey has joined #openstack-oslo | 23:24 | |
| *** stevemar_ has quit IRC | 23:25 | |
| *** stevemar_ has joined #openstack-oslo | 23:26 | |
| *** jecarey has quit IRC | 23:28 | |
| *** sigmavirus24 is now known as sigmavirus24_awa | 23:29 | |
| *** stevemar_ has quit IRC | 23:30 | |
| *** pratikma_ has joined #openstack-oslo | 23:45 | |
| *** salv-orlando has joined #openstack-oslo | 23:45 | |
| *** pratikmallya has quit IRC | 23:49 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!