*** egafford has quit IRC | 00:04 | |
*** Longgeek has joined #openstack-sahara | 00:54 | |
*** Longgeek has quit IRC | 00:58 | |
*** witlessb has quit IRC | 01:06 | |
*** Poornima has joined #openstack-sahara | 01:51 | |
*** hdd has quit IRC | 02:56 | |
*** Krast has quit IRC | 03:13 | |
*** zhidong has quit IRC | 03:41 | |
*** crobertsrh has joined #openstack-sahara | 04:25 | |
*** Networkn3rd has joined #openstack-sahara | 04:26 | |
*** Networkn3rd has quit IRC | 04:27 | |
*** hdd has joined #openstack-sahara | 04:34 | |
*** crobertsrh is now known as _crobertsrh | 04:35 | |
*** hdd_ has joined #openstack-sahara | 04:42 | |
*** hdd has quit IRC | 04:43 | |
*** dmitryme has quit IRC | 04:55 | |
*** dmitryme has joined #openstack-sahara | 04:56 | |
*** tellesnobrega has quit IRC | 05:03 | |
*** tellesnobrega has joined #openstack-sahara | 05:04 | |
*** hdd_ has quit IRC | 05:19 | |
*** Networkn3rd has joined #openstack-sahara | 06:00 | |
*** hdd has joined #openstack-sahara | 06:03 | |
*** hdd has quit IRC | 06:41 | |
*** Networkn3rd has quit IRC | 06:49 | |
*** pcaruana has joined #openstack-sahara | 07:16 | |
*** Poornima has quit IRC | 08:04 | |
*** Poornima has joined #openstack-sahara | 08:18 | |
*** Longgeek has joined #openstack-sahara | 08:33 | |
*** zhidong has joined #openstack-sahara | 08:47 | |
*** tnovacik has joined #openstack-sahara | 08:49 | |
openstackgerrit | Merged stackforge/sahara-ci-config: Add Vanilla 2.6.0 jobs https://review.openstack.org/141762 | 08:51 |
---|---|---|
*** tnovacik has quit IRC | 08:59 | |
*** tnovacik has joined #openstack-sahara | 09:06 | |
*** witlessb has joined #openstack-sahara | 09:07 | |
*** tnovacik has quit IRC | 09:13 | |
*** tnovacik has joined #openstack-sahara | 09:33 | |
*** tnovacik has quit IRC | 10:03 | |
*** skolekonov has joined #openstack-sahara | 10:06 | |
openstackgerrit | Denis Egorenko proposed stackforge/sahara-ci-config: Fix integration_cleanup for Vanilla 2.6 https://review.openstack.org/142413 | 10:19 |
openstackgerrit | Merged stackforge/sahara-ci-config: Fix integration_cleanup for Vanilla 2.6 https://review.openstack.org/142413 | 11:09 |
*** Poornima has quit IRC | 11:50 | |
openstackgerrit | Alexander Ignatov proposed openstack/sahara: Add one more sample for pig job examples https://review.openstack.org/141782 | 13:13 |
*** stannie has joined #openstack-sahara | 13:14 | |
openstackgerrit | Denis Egorenko proposed stackforge/sahara-ci-config: Temporary disable heat engine https://review.openstack.org/142439 | 13:22 |
*** hdd has joined #openstack-sahara | 13:50 | |
*** miqui has joined #openstack-sahara | 14:07 | |
*** tmckay has joined #openstack-sahara | 14:10 | |
*** egafford has joined #openstack-sahara | 14:16 | |
*** _crobertsrh is now known as crobertsrh | 14:19 | |
*** hdd has quit IRC | 14:36 | |
*** pcaruana has quit IRC | 14:51 | |
*** IvanBerezovskiy1 has quit IRC | 14:58 | |
*** IvanBerezovskiy has joined #openstack-sahara | 14:59 | |
*** hdd has joined #openstack-sahara | 15:32 | |
*** tellesnobrega has quit IRC | 15:38 | |
*** miqui_ has joined #openstack-sahara | 15:48 | |
*** skolekonov has quit IRC | 15:58 | |
SergeyLukjanov | folks, we already have a kilo-1 release ;) | 16:00 |
*** Poornima has joined #openstack-sahara | 16:00 | |
*** guo has joined #openstack-sahara | 16:01 | |
*** Networkn3rd has joined #openstack-sahara | 16:02 | |
elmiko | SergeyLukjanov: yay! | 16:05 |
*** guo has quit IRC | 16:07 | |
crobertsrh | :) | 16:07 |
*** Poornima has quit IRC | 17:05 | |
*** hdd has quit IRC | 17:13 | |
*** IvanBerezovskiy has left #openstack-sahara | 17:17 | |
openstackgerrit | Chad Roberts proposed openstack/sahara-specs: Spec for template editing https://review.openstack.org/140448 | 17:38 |
openstackgerrit | Trevor McKay proposed openstack/sahara: Add options supporting DataSource identifiers in job_configs https://review.openstack.org/142248 | 17:53 |
tmckay | So I have a gate unit test that is failing, but the .testrepository logs don't seem to be available for unit tests, or Sahara logs either. I am experimenting with trapping the assertionError myself and embedding debug information in a new AssertionError with re-raise, so that I can see messages to myself in console.html | 17:58 |
tmckay | Anybody else have a good solution for this? | 17:58 |
tmckay | I have a unit test that breaks in CI but does not break locally | 17:58 |
tmckay | don't know how else to make "print()" work, or log(...) | 17:58 |
tmckay | stdout from test code should be available, imho | 18:00 |
tmckay | but if it is, I don't know where ... | 18:00 |
*** Networkn3rd has quit IRC | 18:03 | |
*** Networkn3rd has joined #openstack-sahara | 18:04 | |
openstackgerrit | Trevor McKay proposed openstack/sahara: Add options supporting DataSource identifiers in job_configs https://review.openstack.org/142248 | 18:07 |
elmiko | tmckay: did you get any traction? | 18:18 |
tmckay | well, waiting for CI to run | 18:19 |
tmckay | based on what I did locally, it should tell me what line in the real proxy.py returned True when False was expexted | 18:19 |
elmiko | ok, seems really weird | 18:20 |
tmckay | yeah, it should produce locally | 18:22 |
tmckay | hmm, might have a theory | 18:38 |
elmiko | k | 18:40 |
openstackgerrit | Trevor McKay proposed openstack/sahara: Add options supporting DataSource identifiers in job_configs https://review.openstack.org/142248 | 19:00 |
tmckay | elmiko, I left the return value for a mocked routine unspecified, and it gave a mock value back which was compared using " > 0" | 19:01 |
tmckay | in the debugger, it says "False" | 19:01 |
tmckay | but, what is that based on? | 19:01 |
tmckay | assuming it could just as well be True | 19:02 |
tmckay | mock > 0 is odd at best | 19:02 |
tmckay | I'm surprised it wasn't throwing an exception, actually. Anyway, I fixed that bit up, we'll see what happens | 19:03 |
*** Networkn3rd has quit IRC | 19:04 | |
*** Networkn3rd has joined #openstack-sahara | 19:04 | |
elmiko | tmckay: interesting, not sure i understand fully | 19:04 |
tmckay | elmiko, well, it ended up doing "if count_routine() > 0" in proxy.py, where count_routine had been mocked in the test code. But the return value was left unset | 19:05 |
tmckay | pdb says the return value was a mock object :) | 19:06 |
tmckay | I'm wondering if "mock > 0" is indeterminate, or based on some attribute in the object that can change | 19:06 |
tmckay | so that I could have different results locally vs CI | 19:07 |
tmckay | only smoking gun I could find | 19:07 |
elmiko | huh, good find | 19:09 |
*** hdd has joined #openstack-sahara | 19:10 | |
*** oikawa has quit IRC | 19:20 | |
openstackgerrit | Michael McCune proposed openstack/sahara: Adding database detection to migration tests https://review.openstack.org/142578 | 20:59 |
elmiko | ruhe: i think this will fix the db presence issue ^^ | 21:00 |
*** Networkn3rd has quit IRC | 21:02 | |
openstackgerrit | Merged stackforge/sahara-ci-config: Temporary disable heat engine https://review.openstack.org/142439 | 21:11 |
tmckay | crobertsrh, ping | 21:12 |
crobertsrh | yes? | 21:12 |
tmckay | looking at the template editing spec. I wonder if we can do something simple, like create template.1 automatically if template is edited and referenced | 21:13 |
tmckay | we talked about this a little previously | 21:13 |
tmckay | everything in the UI is name based, everything in the database is id based | 21:13 |
crobertsrh | right...the oslo-versioned-objects stuff might be a good fit for that | 21:14 |
tmckay | if we renamed an existing template with a .X version number, then the new template would show up with the expected name and existing objects in the db would still be linked together by id number | 21:14 |
crobertsrh | are names unique across tenants? | 21:15 |
tmckay | it wasn't clear to me that that was what they were for. | 21:15 |
tmckay | maybe I should read more closely | 21:15 |
crobertsrh | It might not be a perfect fit. Mike mentioned it yesterday and it sounded potentially promising | 21:15 |
tmckay | sounded more like a versioned API kind of thing, to support compatibility over different versions of objects | 21:16 |
crobertsrh | Of course, if we just did our own .X magic, that may work too | 21:16 |
tmckay | it would be simple | 21:16 |
crobertsrh | True...not necessarily the real right fit. Homegrown .X may be better, and potentially simpler | 21:16 |
crobertsrh | So far, the spec is more SPECulation | 21:17 |
tmckay | we would just need some numbering scheme, and maybe decide what to do (or disallow) if someone edited a template.X | 21:17 |
crobertsrh | Hmm...somewhere in my head, this won't work....trying to articulate it now... | 21:18 |
crobertsrh | So...if I have a cluster template C today that references ngt N, then we edit N. The original N becomes N.X and C still references that? That doesn't seem like what the desired behavior is. | 21:19 |
crobertsrh | I was thinking that C should pick-up the changes to N. Only existing clusters would retain their link to N.X. | 21:20 |
crobertsrh | I should probably specify those cases in the spec, whatever behavior we agree on. | 21:20 |
elmiko | crobertsrh: +1 | 21:22 |
tmckay | hmmm. I could imagine that it would chain. If nothing is ultimately referenced by a cluster, we would just change N in place. If N or C was tied to a running cluster, we would have to create C.x and N.x for the existing cluster | 21:22 |
tmckay | and then C and N would be the right thing | 21:23 |
tmckay | the whole trouble is if there is an existing cluster that references C that references N | 21:23 |
tmckay | have to revisit how the links are done | 21:24 |
tmckay | I'm fuzzy at the moment | 21:24 |
crobertsrh | Yeah, my clarity comes and goes | 21:24 |
crobertsrh | I'm pretty sure we do NOT want to affect running clusters. | 21:25 |
tmckay | agreed | 21:25 |
crobertsrh | I'm pretty sure we DO want NGT changes to be picked-up in cluster templates | 21:25 |
tmckay | I'm trying to come up with a way to snapshot existing info tied to a cluster or clusters | 21:25 |
tmckay | also agreed | 21:25 |
crobertsrh | Ok. I'll try to take a look as well, but you are much better versed there than I am. | 21:30 |
tmckay | crobertsrh, in my mind, the fundamental difference between copy and edit is that in the edit case we're doing rename magic so no info that is depended on is destroyed, and the user has objects with the same name but different info when they are done | 21:30 |
tmckay | can it be done? I don't know | 21:30 |
crobertsrh | Heh | 21:30 |
crobertsrh | Maybe a new field for "version" | 21:31 |
crobertsrh | nah, that doesn't work either | 21:31 |
elmiko | we could do something like git and create a previous-ref field that points to a newly created uuid for the older object info, then the updated object can retain the same uuid. | 21:33 |
crobertsrh | Maybe the edited version keeps the original ID, we make a copy of the original object (with new ID) and update those refs in the database? | 21:34 |
elmiko | exactly | 21:34 |
crobertsrh | great minds | 21:34 |
elmiko | i think the edited version has to keep the original id for compatibility | 21:34 |
tmckay | that's what I'm envisioning | 21:34 |
crobertsrh | I think that's the ticket | 21:35 |
tmckay | id doesn't matter so much, it's the names that are troublesome I think | 21:35 |
crobertsrh | Oh, right. | 21:35 |
elmiko | tmckay: i would think id matters for things like endpoints | 21:35 |
crobertsrh | We may need to .X them | 21:35 |
tmckay | I think the originals would preseve their ids and just have their names changed | 21:37 |
tmckay | New stuff would get new ids | 21:37 |
tmckay | that way all the foreign key stuff in the db is preserved | 21:37 |
tmckay | names are unique, but not used as foreign keys I believe | 21:37 |
tmckay | crobertsrh, earlier question, I think every table has a tenant_id, an id, and a name | 21:38 |
elmiko | tmckay: ok, we're on the same page. i thought you were saying that the ids don't matter in terms of updating. | 21:38 |
tmckay | so, effectively, names are unique across tenants. They're all in the same db | 21:39 |
tmckay | elmiko, well, I was thinking of onscreen presentation. Users want the names to be what they expect. You're right, though, way less work if existing links remain untouched | 21:41 |
elmiko | tmckay: totally agreed about names. no question. i just think it's important to have the uuids be the same to reduce client side mess with renamed endpoints | 21:42 |
crobertsrh | So, duplicate uuids are kosher? | 21:43 |
tmckay | you're thinking of URLs where the id is part of the path, and somone has it cached? | 21:43 |
elmiko | crobertsrh: no | 21:43 |
tmckay | no, duplicate ids are impossible | 21:43 |
elmiko | tmckay: something along those lines | 21:43 |
tmckay | yeah, we don't want to break that | 21:44 |
tmckay | In my mind it's simple | 21:44 |
tmckay | 1) old object gets new name | 21:44 |
tmckay | 2) new object gets old name and new id | 21:44 |
tmckay | If you have to bubble up from ngt to cluters, you repeat the operation a level up | 21:44 |
elmiko | ooh, just had a thought | 21:45 |
* crobertsrh is updating the spec on the fly to avoid forgetting any of our wisdom. | 21:45 | |
tmckay | brb | 21:45 |
elmiko | when the original object is updated, and a new "reference" object is created for the previous version the name could be some variation of the original uuid. | 21:46 |
crobertsrh | name.origuuid? | 21:47 |
elmiko | that could work, if we only keep one historical copy | 21:48 |
crobertsrh | we could need multiples | 21:48 |
elmiko | so maybe, name.rev.origuuid or something | 21:49 |
crobertsrh | [original...launch cluster, edit, launch another cluster, edit] would require multiple copies to be kept. | 21:49 |
elmiko | yea totally | 21:50 |
crobertsrh | A little ugly since the user will see those uuids show up in the names in the UI | 21:51 |
elmiko | ooh good point | 21:52 |
elmiko | man... this could get ugly | 21:52 |
crobertsrh | Is the template itself just stored as a blob? | 21:53 |
* crobertsrh could look, but found it easier to ask. | 21:53 | |
* crobertsrh looked. Nope | 21:54 | |
elmiko | are you saying that maybe the template should be stored in the db or something? | 21:54 |
crobertsrh | just thinking about keeping all necessary versions in the same record | 21:55 |
crobertsrh | Or, probably a separate table. | 21:55 |
crobertsrh | Sort of the "microversion" impl | 21:56 |
elmiko | yea | 21:56 |
elmiko | maybe we'll want to have a checkbox to delete the old version when updating to reduce the need for people to manually the delete the old ones? | 21:57 |
crobertsrh | original table would just contain original name, id...and point to the new table that contains all the microversions | 21:57 |
elmiko | new table could definitely work magic for this, assuming no one has an issue with another db migration | 21:58 |
crobertsrh | We should be able to delete any old one that isn't referenced by a running cluster | 21:58 |
elmiko | ok, so old ones might not be a huge issue | 21:58 |
crobertsrh | I don't think so. They could accumulate in a long-running high use system, but they don't really take much room to store. | 21:58 |
elmiko | seemed like keeping track of them might be more the issue | 21:59 |
openstackgerrit | Trevor McKay proposed openstack/sahara: Add options supporting DataSource identifiers in job_configs https://review.openstack.org/142248 | 22:01 |
crobertsrh | I don't think db migrations should be a limiting factor. | 22:02 |
crobertsrh | Or....we could just say that can't edit a template that is referenced by a running cluster. | 22:03 |
crobertsrh | I think things get simple in that case | 22:03 |
crobertsrh | Let them make their own copies explicitly. | 22:04 |
elmiko | i don't think migrations should be an issue either, i just know it's a factor to consider. | 22:04 |
crobertsrh | It would be congruent with not letting users delete templates that are used by a cluster. | 22:04 |
openstackgerrit | Chad Roberts proposed openstack/sahara-specs: Spec for template editing https://review.openstack.org/140448 | 22:10 |
crobertsrh | Is tomorrow's meeting early or late? | 22:11 |
crobertsrh | We should talk about this at the meeting. | 22:11 |
crobertsrh | Ok, I may be talking to myself now. | 22:12 |
elmiko | early | 22:13 |
crobertsrh | Thanks | 22:13 |
elmiko | sorry, got like 4 different channels i'm watching lol | 22:13 |
crobertsrh | I don't think zimbra has a good way to juggle the meetings, does it? | 22:13 |
crobertsrh | heh :) | 22:13 |
openstackgerrit | Trevor McKay proposed openstack/sahara: Add options supporting DataSource identifiers in job_configs https://review.openstack.org/142248 | 22:13 |
crobertsrh | My solution....set both meeting times to show up every week | 22:13 |
elmiko | i dunno, my thunderbird is not syncing with zimbra cal at all. i'm gonna just make 2 entries in my cal for our meetings | 22:13 |
elmiko | lol, nice solution | 22:13 |
crobertsrh | Bonus is that fewer real meetings will be booked with me since I'm "busy" more | 22:14 |
crobertsrh | I always feel like I've been pardoned when a phantom meeting request pops-up | 22:14 |
crobertsrh | I have the same thing for the doc team meetings | 22:14 |
*** crobertsrh is now known as _crobertsrh | 22:15 | |
*** egafford has quit IRC | 23:35 | |
*** hdd has quit IRC | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!