IQSS logo

IRC log for #dataverse, 2019-01-23

Connect via chat.dataverse.org to discuss Dataverse (dataverse.org, an open source web application for sharing, citing, analyzing, and preserving research data) with users and developers.

| Channels | #dataverse index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

All times shown according to UTC.

Time S Nick Message
00:58 pdurbin :)
08:23 jri joined #dataverse
12:04 juancorr22 joined #dataverse
12:14 juancorr22 left #dataverse
12:15 juancorr joined #dataverse
12:56 donsizemore joined #dataverse
13:47 donsizemore @pdurbin morning. another quick question?
13:53 pdurbin donsizemore: hit me
13:53 donsizemore dv-root.json is hard-coded in a normal installation but could be made configurable/template in ansible. should it be?
13:53 donsizemore also, i hope your child is feeling better
13:55 pdurbin She's back at school today. Thanks. She was feeling well enough last night to eat a brownie her sister baked. Man, it's a serious deficiency of Dataverse that you can't edit a small d dataverse via API. Please do whatever you can.
13:56 pdurbin Ah, hard coded here: https://github.com/IQSS/dataverse/blob/v4.10.1/scripts/api/setup-all.sh#L74
13:56 pdurbin I'd call that a bug.
13:56 donsizemore @pdurbin yeah, a grep -r over the source shows a number of places calling it. easy for ansible to make it configurable
13:56 pdurbin I mean, it's fine to have some sane defaults. Keep calling it Root or whatever.
13:57 pdurbin Ok, so you'd have Ansible but a new file in place? In the expected location?
13:58 donsizemore i'd keep the filename but maybe make description and contact configurable
13:58 donsizemore otherwise i'd pull it from github into the playbook and use as-is. may not matter to a developer
14:01 pdurbin Developers are surviving as is. Someone like pameyer would be interested in having it be configurable, I bet.
14:01 pdurbin For phoenix I have a hard-coded custom file: https://github.com/IQSS/dataverse/blob/v4.10.1/scripts/deploy/phoenix.dataverse.org/dv-root.json
14:01 donsizemore also, depending on the datacite credentials passed, i'm telling the "publish root dataverse" task to ignore errors.
14:02 pdurbin Have you considered using the new "FAKE" PID provider?
14:03 pdurbin I'm using it on phoenix: https://github.com/IQSS/dataverse/blob/v4.10.1/scripts/deploy/phoenix.dataverse.org/post#L8
14:03 donsizemore there's a fake pid provider?
14:03 donsizemore will do
14:03 pdurbin Yeah. It does a no-op. Doesn't actually reach out over the Internet on publish.
14:03 pdurbin see https://github.com/IQSS/dataverse/issues/5327
14:04 pdurbin I've been using it from main.yml. Works great. Thanks again for those group vars.
14:06 donsizemore cool. i may tie that in. or you can submit a pull request =)
14:06 pdurbin Well, there's no pull request necessary. Again, I just edit my own copy of main.yml.
14:08 pdurbin If that makes sense.
14:11 donsizemore i remember reading about that. if you're good leaving the role as-is there, i'll X-out the to-do entry i just made =)
14:12 pdurbin I'm not sure what I would change. Not sure what you mean. I'm good, I think.
15:58 pameyer joined #dataverse
15:59 donsizemore joined #dataverse
16:39 donsizemore @pameyer it may interest you to know that i'm now seeing systemd timeouts in vagrant test boxen using dataverse-ansible
16:40 pameyer @donsizemore that is interesting
16:41 pameyer any ideas if there's something else going on when timeouts become more frequent?
16:41 donsizemore @pameyer i bumped it to 240s as an initial test. have to wait a bit to see ;)
16:42 donsizemore @pameyer and my storage can't be frozen, high of 63°F here today
16:43 donsizemore @pameyer service.log shows it starting up and possibly dying
16:44 donsizemore @pameyer what are you setting in limits.conf?
16:45 pameyer @donsizemore I'll check limits.conf; but I don't recall changing anything there
16:45 donsizemore dataverse-ansible needs to do a bit more than it is currently
16:45 donsizemore it's still going to time out. something's in a bad way.
16:46 donsizemore the ansible role pulls a clean vagrant VM each time and downloads the production dvinstall.zip
16:46 donsizemore (and i had hoped to gym over lunch so i may take a break and resume troubleshooting once i'm back)
16:46 pameyer 63F isn't exercise outside weather? ;)
16:47 donsizemore sadly our weightroom has no retractable stadium roof
16:48 pameyer I haven't checked anything from cent7 defaults in limits.conf
16:48 donsizemore glassfish.service: control process exited, code=exited status=143
16:49 pameyer are you starting from a fresh glassfish zip download, or a cached one?
16:49 donsizemore what the guide recommends for solr http://guides.dataverse.org/en/latest/installation/prerequisites.html#solr i do as well for glassfish
16:49 donsizemore fresh
16:49 donsizemore or at least, i'm not caching it anywhere
16:50 donsizemore glassfish doesn't want to stop itself cleanly
16:51 pameyer that reminds me of a case where I changed a asadmin restart to a start/stop sequence
16:51 donsizemore i don't do restarts for this reason
16:51 pameyer that seemed slightly more robust, because I was seeing asadmin restart fall over
16:51 pameyer gotcha - so not that
16:51 donsizemore remember apache's "graceful"?
16:52 donsizemore so a manual glassfish start takes 51s. and it stops itself
16:52 pameyer stops itself meaning no gf process running after a manual start?
16:53 donsizemore meaning no glassfish process after a manual stop
16:53 pameyer sorry - was confused
16:53 donsizemore my statement was ambiguous
16:54 pameyer if I'm interpreting things correctly, 143 translates to improperly handled SIGTERM
16:54 donsizemore and glassfish starts back up cleanly using systemd. it's something in the install process as you found
16:55 donsizemore i never did know when to quit
16:55 pameyer ;)
16:55 pameyer do you notice similar oddness with interactive vs non-interactive that I was seeing?
16:56 donsizemore i've only started to poke at it
16:56 donsizemore Unprocessed event : UnprocessedChangeEvent{PropertyName=jvm-options, OldValue = [-XX:MaxPermSize=192m,
16:56 donsizemore reason = Removal of: -XX:MaxPermSize=192m cannot take effect without server restart, when = 1548261027729]]
16:57 donsizemore actually, it's pissed about every jvm-option
16:57 donsizemore then it logs "shutdown initiated"
16:59 donsizemore Exception while visiting META-INF/versions/9/org/junit/platf​orm/commons/util/ModuleUtils.class of size 10770 java.lang.IllegalArgumentException
17:00 donsizemore ima kill it all and start from scratch
17:00 pameyer that junit exception has me curious
17:01 donsizemore there were 3 or 4 of them.
17:01 donsizemore will you be upset if i head to the gym? then resume banging my head on this
17:01 pameyer they may not be meaningful
17:01 pameyer not at all
17:01 donsizemore i know, but i feel better when i go to the gym
17:02 pameyer I'll poke things a bit while doing an upgrade
17:15 pameyer IllegalArgumentException from junit doesn't appear to be a direct culprit (although I'd be surprised if those classes were intended to be deployed)
17:20 pameyer I'm not seeing anything for jvm-options other than the usual "restart me" ones
17:24 pdurbin pameyer: I'm coming at you to discuss https://github.com/IQSS/dataverse/issues/5424#issuecomment-456890746 on my way to DataFest. :)
17:24 pameyer @pdurbin thanks for the heads up
17:40 jri joined #dataverse
18:09 donsizemore joined #dataverse
18:10 donsizemore @pameyer this may be a goose herring, but i'm back now and want to review the recent security changes to systemd
18:23 pameyer @donsizemore sounds reasonable.  I'm about to head to a talk; should be back later
18:24 jri joined #dataverse
18:25 pameyer the glassfish stoppages I was seeing were in docker, and those weren't using systemd at all - so very likely unrelated
18:34 donsizemore @pameyer agreed, but it's where i'm breaking down. just tinkered with my limits.conf settings and re-running
18:39 pdurbin go don go
18:41 pdurbin pameyer: I'm digging the diagram she showed: https://quantdare.com/what-is-the-difference-between-bagging-and-boosting/
18:48 donsizemore updating limits.conf fixed things, at least for me
18:48 donsizemore (duplicating the solr nofile/nproc entries for glassfish)
18:49 donsizemore now to see if that's true the 2nd time
19:13 Jim__ joined #dataverse
19:33 donsizemore yis =)
19:56 pameyer pdurbin: I was surprised it was possible to give a boosting talk without mentioning viola-jones ;)
19:56 pameyer @donsizemore good you've got it fixed :)
19:57 pameyer and sounds like I should see if limits.conf helps with docker when I get back to that
19:57 donsizemore @pameyer i hope so. i don't like when things don't work and there's no smoking gun
19:58 pameyer me either - glassfish crashing with nothing in any logs seems like one of those "should never happen" things
19:58 pameyer but it does happen
20:43 donsizemore @pameyer so, i'm at 5 runs now with two different vagrant source VMs — 3 succeed, 2 fail. ansible sez it's editing limits.conf but it's not. i'm heading home
21:32 pdurbin pameyer: her slides are up already: https://slides.com/dbouquin/datafest2019_sgbm#/

| Channels | #dataverse index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

Connect via chat.dataverse.org to discuss Dataverse (dataverse.org, an open source web application for sharing, citing, analyzing, and preserving research data) with users and developers.