Log of the #duraspace-ff channel on chat.freenode.net

Using timezone: Eastern Standard Time
* github-ff joins09:09
[fcrepo4] mikedurbin opened pull request #279: Added support for version reversion. (master...revert-version) http://git.io/F0uTSw
* github-ff leaves
<pivotal-bot>Mike Durbin added comment: "https://github.com/futures/fcrepo4/pull/279" https://www.pivotaltracker.com/story/show/60883346
Mike Durbin finished "Version Management" https://www.pivotaltracker.com/story/show/6088334609:10
Mike Durbin added "Enable Version Deletion" https://www.pivotaltracker.com/story/show/6917756009:11
Mike Durbin started "Specify required internal queries" https://www.pivotaltracker.com/story/show/68987038
* edInCo joins09:15
* mikeAtUVa leaves09:20
<pivotal-bot>Andrew Woods added "Determine potential of horizontal scaling" https://www.pivotaltracker.com/story/show/6917885209:28
Andrew Woods started "Determine potential of horizontal scaling" https://www.pivotaltracker.com/story/show/69178852
* mikeAtUVa joins
<pivotal-bot>Andrew Woods edited "Default WAR is larger than one-click WAR" https://www.pivotaltracker.com/story/show/6900167609:29
Andrew Woods edited "Test 2-Node Ingest on UCSD cluster" https://www.pivotaltracker.com/story/show/61835946
Andrew Woods edited "Enable Version Deletion" https://www.pivotaltracker.com/story/show/69177560
Andrew Woods edited "Enable Version Deletion" https://www.pivotaltracker.com/story/show/6917756009:30
* ksclarke joins09:31
<pivotal-bot>Esme Cowles started "Enable index rebuild" https://www.pivotaltracker.com/story/show/65168574
Andrew Woods edited "FedoraFileSystemConnectorIT test fails when running mvn install without running a clean first" https://www.pivotaltracker.com/story/show/6901380210:20
Andrew Woods edited "RDF-ify serialization formats" https://www.pivotaltracker.com/story/show/6522140410:21
* scossu joins10:31
<pivotal-bot>Esme Cowles added comment: "I must have had fcrepo4 running on port 8080 when I ran the ITs, so the indexer was able to retrieve records..." https://www.pivotaltracker.com/story/show/6516857410:48
<awoods>mikeAtUVa ?11:00
* longshou joins11:08
<escowles>longshou: i'm going to get some lunch right now, but we can look at setting up the repo/indexer/triplestore on the VMs in 30 minutes?11:31
<longshou>Sounds good. Thanks a lot.11:36
<pivotal-bot>Andrew Woods added "F4 Test Dataset" https://www.pivotaltracker.com/story/show/6919058811:42
Andrew Woods added "Document rebuild-index capability" https://www.pivotaltracker.com/story/show/6919106611:47
Andrew Woods edited "Document rebuild-index capability" https://www.pivotaltracker.com/story/show/69191066
Andrew Woods edited "Document rebuild-index capability" https://www.pivotaltracker.com/story/show/69191066
Esme Cowles added "Configure indexer webapp to run fcrepo4 for integration tests" https://www.pivotaltracker.com/story/show/6919228012:01
<escowles>longshou: i'm back -- let me know when you're ready12:04
<longshou>Esme, let's start then.12:30
<escowles>longshou: ok, i think the first thing is getting fuseki running12:32
i see that on lib-devsandbox1 and was able to run it without problems
<longshou>Yes. I installed it on both lib-devsandbox1 or lib-devsandbox4.12:34
<escowles>ok, i'll stop my instance on lib-devsandbox1 so you can start it and be able to see the output12:35
<longshou>I started it on lib-devsandbox1. Would you like to keep it running?12:39
<escowles>also, one quirk of the VMs is they are generally firewalled so you can't just connect to random ports (like 3030 for fuseki), but you can tunnel the port using ssh: ssh -f lib-devsandbox1.ucsd.edu -L 3030:lib-devsandbox1.ucsd.edu:3030 -N
yes, please keep fuseki running while we get the repo and indexer setup on another machine
the next thing to do is compile fcrepo4 and the indexer -- i'm compiling fcrepo4 on lib-devsandbox3 right now -- can you go ahead and compile fcrepo4 on lib-devsandbox2?12:42
<longshou>Sure.
<escowles>longshou: i got a test failure in FedoraHtmlResponsesIT, but i just noticed the fcrepo4 project hadn't been updated from github. so i pulled the most recent changes and i'm starting again12:49
<longshou>I'm restarting the build now.12:54
I got up-to-date message for a pull. Have you update it on all other VMs?12:56
<escowles>i'm not sure -- i may have updated it on lib-devsandbox2 when i was testing the build issue yesterday12:58
after updating, the build finished successfully for me13:01
* gregjansen joins13:04
<longshou>Esme, it's still running on lib-devsandbox2. Maybe we can start to build the jms indexer on lib-devsandbox4 now?13:05
<escowles>i think we should build and run the indexer on the same machine as fcrepo4 -- but with a setting to point it to fuseki on lib-devsandbox113:06
<longshou>The build finished successfully on lib-devsandbox2 now.13:07
<escowles>a more complicated setup would be to have the indexer on the same machine as fuseki, with updated config to listen to two fcrepo4 instances on other machines
ok, next is building the indexer -- do you have it cloned on lib-devsandbox2?13:08
* gregjansen leaves
<longshou>Do we need any config changes before we build the indexer?
* gregjansen joins
<escowles>no -- there are system properties we can set to change the repo and fuseki hosts, so those can be set at run time13:09
<longshou>Then let's build it on lib-devsandbox1. The codes are there so just go ahead to build it so that you can see the errors too.13:11
<escowles>i'm building the indexer on lib-devsandbox3 right now
it's stopped at "(SparqlIndexerIT) Discovered 0 triples, waiting for 3..." for a few minutes now
<longshou>I ran into the hung too yesterday when I built it locally in my Mac with the test assert count corrected to 2 or > 0.13:14
<escowles>it just finished with a failure: "SparqlIndexerIT.indexerTest:80 Triple should return from search! expected:<1> but was:<2>"13:15
i'm going to change SparqlIndexerIT.java:80 to: assertTrue("Triple should return from search!", countQueryTriples(sparqlQuery) > 0);13:16
<awoods>escowles/longshou: fcrepo-jms-indexer-pluggable builds on my local machine, but it has been hanging on Jenkins: http://ci.fcrepo.org/jenkins/view/FF/job/fcrepo-jms-indexer-pluggable/13:17
<longshou>I saw it the same yesterday locally on my Mac and the VMs.
<awoods>all of the "gray" builds are killed hangs.
<pivotal-bot>Mike Durbin added comment: "The querying is sufficient... now we just need locking.13:18
I updated the wiki page to reflect this." https://www.pivotaltracker.com/story/show/68987038
Mike Durbin finished "Specify required internal queries" https://www.pivotaltracker.com/story/show/68987038
Mike Durbin started "Add node-level or graph-level locking." https://www.pivotaltracker.com/story/show/66093788
<escowles>awoods: i've seen some issues with it taking a long time for the events to be processed and show up (it waits 15 seconds for some of them, and i saw them take 12 seconds)
longshou: with that change, the indexer now builds for me13:20
<pivotal-bot>Andrew Woods added "Investigate hanging fcrepo-jms-indexer-pluggable build hangs" https://www.pivotaltracker.com/story/show/69198694
Andrew Woods edited "Investigate hanging fcrepo-jms-indexer-pluggable build hangs" https://www.pivotaltracker.com/story/show/69198694
Andrew Woods accepted "Specify required internal queries" https://www.pivotaltracker.com/story/show/68987038
* github-ff joins13:23
[fcrepo-jms-indexer-pluggable] escowles created hangfix (+1 new commit): http://git.io/mBKXdg
fcrepo-jms-indexer-pluggable/hangfix 8000795 Esmé Cowles: Updating test to accept more than the expected number of triples
* github-ff leaves
<escowles>longshou: i've pushed that test fix to a new branch called hangfix -- can you checkout that branch and try building the indexer again?13:24
<longshou>I am building it locally on my Mac now. Maybe readme need to update for the right instruction to build the indexer first too.13:32
<escowles>yes, there are definitely some issues around building the indexer and the dependency on fcrepo4 that need to be smoothed out13:35
* gregjansen leaves13:39
<awoods>escowles: I am thinking the build-dependency is an issue that maven can resolve by picking up the fcrepo4 snapshot dependency instead of requiring folks to build fcrepo4 locally in order to build jms-indexer: https://www.pivotaltracker.com/story/show/6914805213:49
<pivotal-bot>feature: Pick-up fcrepo4 snapshot dependencies in fcrepo-jms-indexer-pluggable (unstarted) / owner:
<escowles>awoods: that would be much more convenient than having to compile fcrepo4 locally first13:50
<awoods>escowles: jenkins pushes snapshot builds to sonatype: https://oss.sonatype.org/index.html13:52
escowles: we just need to add that repo to the top-level pom.xml, enabled for snapshots.
<longshou>Esme, It looks good. The build finished successfully n my Mac. To start the indexer, we just need to start the indexer with "mvn -D jetty.port=9999 -D fuseki.host=lib-devsandbox3 jetty:run" since we'll start fcrepo4 on lib-devsandbox3 along with the indexer, right?13:54
<escowles>longshou: yes, that looks right13:55
<longshou>Should we start fcrepo before starting the indexer or the order to strat the apps doesn't matter? Are there any issues with the random 3030 port connecting among the VMs?14:00
* scossu leaves14:02
<escowles>you don't need to restart fcrepo4 -- it and fuseki should be started before the indexer14:04
and there won't be any problem connecting between the VMs on different ports -- they all have unrestricted access to each other. it's just outside connections that are restricted14:05
<longshou>I see. I just started fcrepo fcrepo-webapp-4.0.0-alpha-5-SNAPSHOT-jetty-console.war on lib-devsandbox2. Have you start it on lib-devsandbox3?14:10
<escowles>no, i haven't started it yet -- i'll start it now14:11
ok, i've started fcrepo4 on lib-devsandbox3 and i started the indexer with this command line:14:15
mvn -D jetty.port=9999 -D fuseki.host=lib-devsandbox1 jetty:run14:16
longshou: i've created an indexable object and can see the triples synced to fuseki14:20
* gregjansen joins14:23
<longshou>It sounds good. I am able to create new objects through the RES API on lib-devsandbox2 now. Thank you very much for your help on setting it up.14:25
* gregjansen leaves14:27
<escowles>longshou: cool -- let me know if you have any more problems14:28
* scossu joins14:32
<pivotal-bot>Esme Cowles started "Document rebuild-index capability" https://www.pivotaltracker.com/story/show/6919106614:39
Esme Cowles added comment: "This works when manually tested. Adding configuration for running fcrepo4 (#69192280) should fix the IT." https://www.pivotaltracker.com/story/show/6516857414:40
Esme Cowles unstarted "Enable index rebuild" https://www.pivotaltracker.com/story/show/65168574
* scossu leaves14:41
<longshou>Esme, is there a way to examine the contents in fuseki through http://lib-devsandbox1.ucsd.edu:3030/control-panel.tpl with the firewall issue?14:42
<escowles>longshou: yes, you can tunnel a port on your local machine to lib-devsandbox1 using ssh: ssh -f lib-devsandbox1.ucsd.edu -L 3030:lib-devsandbox1.ucsd.edu:3030 -N14:43
then you can go to http://localhost:3030/ in your browser and it will be tunneled to http://lib-devsandbox1.ucsd.edu:3030
<longshou>Cool! It works, Esme. Thank you very much.14:48
* gregjansen joins15:03
* scossu joins15:04
* longshou leaves15:07
* gregjansen leaves15:13
<pivotal-bot>Esme Cowles added comment: "I have updated the wiki page on indexer configuration with examples of using curl to reindex content: ""15:24
https..." https://www.pivotaltracker.com/story/show/69191066
Esme Cowles finished "Document rebuild-index capability" https://www.pivotaltracker.com/story/show/69191066
* gregjansen joins15:36
<pivotal-bot>Esme Cowles added comment: "I think the issue I was having was related to parallel builds. I had set an option to use multiple threads...." https://www.pivotaltracker.com/story/show/6868919816:47
* mikeAtUVa leaves16:57
<pivotal-bot>Andrew Woods added comment: "@escowles, Jenkins was doing "-T 3C", but that flag has since been removed. Could you create a branch named..." https://www.pivotaltracker.com/story/show/6868919817:10
* gregjansen leaves17:20
* longshou joins17:46
* github-ff joins18:01
[fcrepo4] escowles created federationTests2 (+1 new commit): http://git.io/M5PGnA
fcrepo4/federationTests2 7811dc5 Esmé Cowles: Re-enabling filesystem federation copy tests
* github-ff leaves
<pivotal-bot>Esme Cowles added comment: "@awoods I've just pushed fcrepo4/federationTests2 to re-enabled the filesystem federation tests" https://www.pivotaltracker.com/story/show/6868919818:02
<ksclarke>would the mockito tests in fcrepo-http-api be the place to learn about mocking fedora restful responses?
<awoods>escowles: there is a trigger for a branch named "projection-tests"18:03
hence the pivotal comment.
<escowles>awoods: ok, i'll rename it
* edInCo leaves18:04
* github-ff joins
[fcrepo4] escowles created projection-tests from federationTests2 (+0 new commits): http://git.io/6LHLrw
* github-ff leaves
<awoods>ksclarke: you are looking for unit tests? or integration tests?18:05
<ksclarke>well, I've gone down a bit of a rabbit hole with the benchtool -- it does use a thread pool, but it doesn't set the httpclient thread count to match the executor thread count so I was going to write a unit test for a change I was making18:06
it's separate from fcrepo4 so I thought an integration test would be too difficult to wire in (?)18:07
so was looking at unit test with mock fedora responses for the things benchtool wants to do
<longshou>Esme, it looks like that the single fuseki/indexer system may not work as our expectation. After restarting fuseki, fcrepo on two VMs, and the jms indexer in order, the data ingested into fuseki is lost and I see the ingexer is still working but no longer ingest the triples into fuseki. Are you able to replicate it?18:08
<ksclarke>I'm not really testing those things though so my other option was just to ignore logging when it complains it can't connect and just have my test count the threads in the pool
<awoods>ksclarke: there are many unit test examples around. I am not sure exactly what response mocking you want to do, but fcrepo-http-api is a good place to look for inspiration.18:09
<ksclarke>(basically was just looking to test that -t # set thread count in the httpclient and the executor)
<awoods>escowles ^^18:10
<ksclarke>the way benchtool is wired, unless I just ignore the log complaints, I'd need to mock the ingest and purge actions that it performs18:11
but think this is a bit of a rabbit hole and out of the way of the testing ticket
so not sure it makes a lot of sense to sink a bunch of time into it
<awoods>ksclarke: I trust your judgment18:12
<escowles>i'm not a huge fan of making a pile of mocks to test simple things like that...
<awoods>longshou: did you use the "./fuseki-server --mem" option?
<ksclarke>yeah, seems like a lot of work for very little payoff/usefulness
<awoods>longshou: in which case the in-memory triplestore would be flushed on restart.18:13
all: I need to step away.
* ksclarke had time to kill while the larger performance tests were running today
<longshou>Yes. I used the "./fuseki-server --mem" option.18:15
But it's interesting that the indexer no longer ingest triples into it though it's retrieving events.18:18
<escowles>ksclarke: i think this isn't what you want to do, but i did come across a good example of mocking an http response just yesterday: https://github.com/futures/fcrepo-jms-indexer-pluggable/blob/reindex/fcrepo-jms-indexer-core/src/test/java/org/fcrepo/indexer/IndexerGroupTest.java#L17018:21
<ksclarke>escowles: interesting, thanks18:24
* travis-ci joins
[travis-ci] futures/fcrepo4#1693 (federationTests2 - 7811dc5 : Esmé Cowles): The build passed.
[travis-ci] Change view : https://github.com/futures/fcrepo4/commit/7811dc5c40b5
[travis-ci] Build details : http://travis-ci.org/futures/fcrepo4/builds/22647562
* travis-ci leaves
* travis-ci joins18:27
[travis-ci] futures/fcrepo4#1694 (projection-tests - 7811dc5 : Esmé Cowles): The build passed.
[travis-ci] Change view : https://github.com/futures/fcrepo4/compare/projection-tests
[travis-ci] Build details : http://travis-ci.org/futures/fcrepo4/builds/22647771
* travis-ci leaves
* ksclarke leaves18:50
<awoods>escowles: It looks like Jenkins is still unhappy: http://ci.fcrepo.org/jenkins/view/FF/job/fcrepo4-experimental/3/19:17
* longshou leaves19:22
<pivotal-bot>Longshou Situ added comment: "Do we have any ways to configure fcrepo to post the events to a specific ActiveMQ host or not? It looks li..." https://www.pivotaltracker.com/story/show/6885335419:25
* scossu leaves19:28
* longshou joins19:48
<pivotal-bot>Andrew Woods added comment: "Maybe I do not understand your question exactly, but it seems that the "brokerURL" of the "connectionFactor..." https://www.pivotaltracker.com/story/show/6885335420:00
Longshou Situ added comment: "Yes. I think we need to configure it this way in a multiple fcrepo instances with one jms indexer and a tr..." https://www.pivotaltracker.com/story/show/6885335420:06
Andrew Woods added comment: "Sounds good. Feel free to add the ${jms.host:localhost} update to jms.xml." https://www.pivotaltracker.com/story/show/6885335420:12
* longshou leaves20:32
* escowles_ joins21:18
* awoods leaves21:22
* escowles leaves
* awoods joins21:24
<pivotal-bot>Andrew Woods added comment: "Pending revision based on code review comments." https://www.pivotaltracker.com/story/show/6492915622:19
Andrew Woods rejected "Support for authentication in JMS indexer" https://www.pivotaltracker.com/story/show/64929156