Evergreen ILS Website

IRC log for #evergreen, 2021-11-10

| Channels | #evergreen index | Today | | Search | Google Search | Plain-Text | summary | Join Webchat

All times shown according to the server's local time.

Time Nick Message
06:01 pinesol News from qatests: Testing Success <http://testing.evergreen-ils.org/~live>
07:24 collum joined #evergreen
07:28 rjackson_isl_hom joined #evergreen
08:29 mmorgan joined #evergreen
08:36 mantis joined #evergreen
09:04 Dyrcona joined #evergreen
10:12 rfrasur joined #evergreen
11:29 jvwoolf joined #evergreen
11:59 Christineb joined #evergreen
12:27 collum joined #evergreen
13:08 Dyrcona joined #evergreen
13:27 khuckins joined #evergreen
13:39 csharp_ re: systemd, see bug 1912796
13:39 pinesol Launchpad bug 1912796 in OpenSRF "Docs: add systemd setup to README" [Wishlist,New] https://launchpad.net/bugs/1912796
13:40 csharp_ erg - I was thinking of websocketd - nevermind
13:42 Dyrcona You could add a systemd init file for websocketd, though. That would be helpful.
13:43 Dyrcona So, I'm looking into bug 1937294, and it looks like part of the problem is that all of concerto doesn't load.
13:43 pinesol Launchpad bug 1937294 in Evergreen "Updating Evergreen for Newer PostgreSQL Versions" [Undecided,New] https://launchpad.net/bugs/1937294 - Assigned to Jason Stephenson (jstephenson)
13:43 Dyrcona Copy numbers are different for one thing, but the db server didn't log any errors that looked relevant.
13:45 JBoyer I have separate unit files for opensrf, sip, z3950, clark, and websocketd. Haven't been able to make myself genericize them and get them in a proper branch though. :(
13:48 sandbergja joined #evergreen
13:49 sandbergja Hi all -- I want to talk about releases and release notes.  I'm doing so here (rather than #evergreen-release) so that it is logged.
13:50 sandbergja Due to added pressures at work, I've stopped regularly checking #evergreen-release.  It just hasn't seemed like a good use of my brain space to monitor it constantly just in case a release is planned, which only happens every month or two.
13:50 sandbergja I am still happy to be pinged over email if release notes need to be compiled.
13:51 sandbergja Or if we ever get back to planning releases via google calendar, I can use that to block off time to do point release notes
13:53 sandbergja I think it would be good to re-evaluate how useful having the separate #evergreen-release channel is -- I've heard it described as a barrier to getting more help with releases
13:53 sandbergja But that's a longer term discussion :-)
13:54 sandbergja short term tl;dr: if you need some release notes from me, send me an email, as I am not monitoring #evergreen-release regularly :-)
13:55 Dyrcona sandbergja: Fair enough. We're still using the Google sheet and releases are on Google calendar, though the dates are subject to change without notice.
13:55 sandbergja also a huge thank you to abneiman for stepping in to do release notes for the past few releases -- they had tons of bug fixes, which makes it a really big lift
13:55 sandbergja abneiman++
13:55 Dyrcona I think most of us have been busy lately.
13:56 Dyrcona abneiman++ indeed
13:56 sandbergja Dyrcona: true -- I forgot about the spreadsheet
13:57 sandbergja but that becomes yet another place to check
13:59 JBoyer sandbergja, I think a lot of the recent (since the hackaway at least) discussion has been about making release notes handling more automatic so we don't need to have someone sign up to explicitly take that task and handle it, with the hope that whoever may be putting that release together can just run ./build/tools/make_relnotes.sh or whatever as part of their process.
13:59 JBoyer So hopefully there would need to be fewer people involved in putting a release together.
14:00 JBoyer (I think; I couldn't really participate in the hackaway as much as I intended to)
14:00 sandbergja I think that would be really helpful!
14:01 sandbergja a lot of the problem boils down to: how do you coordinate 3-4 super-busy people
14:01 sandbergja and even if it turns into 2-3 super-busy people to coordinate, it's still an improvement
14:02 sandbergja (I also couldn't participate in the hackaway as much as I intended, so I'm just guessing based on what I heard)
14:05 Dyrcona Well, our release process ought to be a lot simpler, so it doesn't require 3 or more people.
14:10 sandbergja hear hear
14:33 Dyrcona I'm seeing some weird stuff when loading concerto on Pg 11+. It looks like the parallel request workers could be changing IDs and possibly causing some updates not to happen.
14:35 Dyrcona IDK, the more I look, the weirder things get.
15:38 Dyrcona Maybe we should just switch to MariaDB. *ducks*
15:42 rhamby bad Dyrcona!
15:43 Dyrcona Well, PostgreSQL is breaking our data.
15:43 Dyrcona I'm getting very different results when loading the sample data on Pg 11+.
15:43 rhamby ug
15:43 rhamby no bueno
15:44 Dyrcona I'm trying to disable the parallel workers, but so far it's not working.
15:44 rhamby actually my opinion on MariaDb is mostly positive though it lakes a few things that really irk me like inheritance
15:44 rhamby s/lakes/lacks/
15:46 Dyrcona rhamby: Does inheritance irk you or does MariaDB's lack of inheritance irk you?
15:47 rhamby Dyrcona the lack of it, I use it a lot for data work in postgres
15:47 * Dyrcona hasn't actually used MariaDB. I've stuck mainly with MySQL despite it being acquired by the Evil Empire.
15:48 jeff Dyrcona: can you share steps to reproduce the issue you're seeing, and the symptoms? I'm curious.
15:48 Dyrcona OK. Thought I'd ask, because a lot of people don't like table inheritance.
15:48 Dyrcona jeff: Just --load-all-sample on Pg11+ and watch the tests fail, particularly, live_t/20-hold-targeter.t
15:49 Dyrcona I'm comparing between Pg10 and Pg11 side by side and they get different numbers of copies in different copy locations.
15:50 Dyrcona I'm assuming parallel execution is the problem, but it could be something else. Unfortunately, there are no errors or other useful messages logged.
15:50 rhamby just a straight up count(*) from asset.copy where location = foo; ?
15:50 Dyrcona So far, trying to disable parallel execution in the SQL to load the data isn't working.
15:50 Dyrcona Hang on.
15:53 Dyrcona Since Evergreen paste is busted, I'll paste a one-liner here:
15:53 Dyrcona select copy_location.id, copy_location.name, copy_location.holdable from asset.copy_location join asset.copy on copy.location = copy_location.id join asset.call_number on copy.call_number = call_number.id and call_number.record = 42 order by id;
15:54 Dyrcona On Pg 11 you get 10 copies in non-holdable locations. On Pg 10, I get 7 copies in non-holdable copy locations. Also the location names are different.
15:55 Dyrcona Also, if you do 'request open-ils.hold-targeter open-ils.hold-targeter.target {"hold":263}' in srfsh, you get different results. This is from the failing hold-targeter test and record 42 is the record in question.
15:55 Dyrcona I should also mention that the copies and call numbers tend to get different ids in either Pg version.
15:57 Dyrcona I'm going to bug the good people of #postgresql about disabling parallel workers while running a script.
16:01 Dyrcona I'm also starting to think that the problem isn't parallel execution per se.
16:17 jeff I suspect it's "test data load relies on assumptions about sequences that are no longer valid", but I haven't gotten hands-on with it yet. Might carve out time tonight.
16:18 Dyrcona jeff: It's more than that.
16:23 jeff ah, I see.
16:29 Dyrcona That query I pasted gives very different results, so I think it is more than just assumptions about ids. Setting max_parallel_workers_per_gather to 0 doesn't help, so I'm starting to think the problem is something else.
16:33 Dyrcona On the plus side, psql interactive mode in Emacs makes it easy to compare the results in Pg10 vs. Pg11.
16:46 Dyrcona There is such a thing as being too clever.
17:27 mmorgan left #evergreen
17:28 abneiman belatedly mmorgan++ for also release-notes-ing for those two monster releases! & pursuant to what sandbergja said above, I'm always willing to help but don'tspend a lot of time in IRC (and windows doesn't like to give me Quassel notificaitons), so, alternate methods of flagging me down are always OK :)
18:01 pinesol News from qatests: Testing Success <http://testing.evergreen-ils.org/~live>
19:43 Christineb joined #evergreen
19:43 troy joined #evergreen
19:51 troy joined #evergreen
19:52 Christineb joined #evergreen
20:01 ejk joined #evergreen
20:11 eby joined #evergreen
20:24 degraafk joined #evergreen
20:24 jeffdavis joined #evergreen
21:15 jvwoolf joined #evergreen

| Channels | #evergreen index | Today | | Search | Google Search | Plain-Text | summary | Join Webchat