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

Using timezone: Eastern Standard Time
* dwilcox joins08:30
* gregjansen joins09:17
Has anyone else seen 410 Gone responses when running benchtool against a load balancer? It works fine against any individual node.09:18
<awoods>gregjansen: no... but of course I do not have a cluster up.09:26
<gregjansen>Strange, it seems to work fine via the web UI on the LB.09:31
just not through benchtool
* kaarefc joins09:32
* ajs6f leaves09:35
* kaarefc leaves09:36
* dwilcox leaves09:44
* dwilcox joins
* dwilcox_ joins09:47
<pivotal-bot____>Mike Daines added "Refactor authz to provide more information to PEPs and make API clearer for implementers" https://www.pivotaltracker.com/story/show/6694612209:48
* dwilcox leaves
* mikeAtUVa joins
<gregjansen>Okay awoods: the issue with 410 Gone was due to transactions being used in the setup stage of the benchtool. After turning that off with Ben's help, it worked fine. Of course this means that I cannot measure the transaction performance on my cluster. Sounds like we need a LB config that is sticky w/respect to transactions. Anybody have experience with that sort of thing??09:50
<benpennell>and yes, tx's are enabled by default for setup/teardown phases in benchtool (for supporting fedora versions) presently because they provide a sizable performance boost. but we could either change that default or change the default for cluster tests10:00
* scossu joins10:02
* ksclarke joins10:05
* kaarefc joins10:20
* kaarefc leaves10:27
* nbanks joins10:33
* ajs6f joins10:54
Hangout in five: https://plus.google.com/hangouts/_/event/c1glu6soq43r1rr6ou17qtobug810:55
<scossu>ajs6f: re: failing tests: Actually, there are some earlier tests that fail (in versioning HTTP API) due to restrictions:11:11
javax.jcr.RepositoryException: No valid property definition on node '/testRepositoryWideAutoVersioning' with primary type 'fedora:baseObject' and mixin types [fedora:resource,fedora:object,fedora:fileContainer] for the property: jcr:versionHistory="87a0a8c317f1e77ee77e00-845f-4a84-8a7b-fe7c1b9e0eb2"
ajs6f: Didn't you say that jcr:versionHistory is not persisted?
<fasseg>ajs6f: I think it's them: http://www.uvt.ro/ro/11:12
ajs6f: the ticket is https://www.pivotaltracker.com/story/show/6651423611:13
<pivotal-bot____>feature: Add SPARQL actions to bench-tool (unscheduled) / owner:
Kevin Clarke added comment: "Yes, we should no longer need the separate fcrepo checkstyle plugin once the new upstream release is made. ..." https://www.pivotaltracker.com/story/show/6614493011:14
<ajs6f>fasseg: thnx11:16
scossu: jcr:versionHistory _should not_ be persisted. You may have discovered a subtle bug. Or, the tests may be wrong— IOW, they may be trying to persist things they shouldn't be able to persist. Can you take a quick look at those test and see if they are (incorrectly) trying to persist that property?11:17
fasseg: could it be : http://www.upt.ro/english/11:18
?
<scossu>ajs6f: ok. Give me a few mins.11:19
<ajs6f>scossu: Thank you very much!
<scossu>BTW, for some reason, I had to change all the HTTP createObject methods to use a random UUID for the object ID, instead of a literal one. Apparently, these methods are called multiple times and the subsequent times createObject is trying to create a node under it. This fails because the node type is restrictive.11:20
This didn't happen in previous tests, so I wonder if anything changed in the HTTP create methods in the tests or in the upstream methods.11:21
* nbanks leaves11:28
<ajs6f>scossu: No, I think you discovered a weakness in the tests.11:33
You're doing the right thing by using a UUID.
We shoul dhave been doing that all along.
<pivotal-bot____>Gregory Jansen added "Support sticky load balancer routes for transactions" https://www.pivotaltracker.com/story/show/6695500211:34
<scossu>ajs6f: I did a 'grep -IRn versionHistory . --exclude-dir=target' and nothing showed up that seems to assign that property explicitly.11:38
<ajs6f>scossu: Right. But I'm thinking that what might be happening in a test is: "Retrieve an RDF description, alter it, try to persist it" and the predicate in question isn't being filtered out, though it should be.11:39
daines: What is the ticket related to the change to avoid requiring a servlet container?11:47
<daines>ajs6f: https://www.pivotaltracker.com/story/show/66946122 (this is the ticket for changes to authz, the proposal on the wiki page was changed to take into account the possibility of not having a servlet container present)11:49
<pivotal-bot____>feature: Refactor authz to provide more information to PEPs and make API clearer for implementers (unscheduled) / owner:
<ajs6f>daine: Thanks!
<pivotal-bot____>Mike Daines estimated "Refactor authz to provide more information to PEPs and make API clearer for implementers" as 3 points https://www.pivotaltracker.com/story/show/66946122
Mike Daines started "Refactor authz to provide more information to PEPs and make API clearer for implementers" https://www.pivotaltracker.com/story/show/66946122
* tecoripa joins12:01
* tecoripa leaves
* github-ff joins12:34
[fcrepo4] kaisternad opened pull request #262: Remove unused code from SessionFactory (master...fcrepo-65933082) http://git.io/aaXDXg
* github-ff leaves
<pivotal-bot____>Kai Sternad finished "Remove dead SessionFactory code" https://www.pivotaltracker.com/story/show/65933082
Kai Sternad started "PATCH not working in Transactions" https://www.pivotaltracker.com/story/show/6626285612:45
<scossu>ajs6f: I fixed the versioning problem. The code is on my fork if you want to cherry-pick.
Now it's back to the previous failure with the SPARQL test…12:46
<ajs6f>scossu: That's great— can you say roughly what it turned out to be?12:48
<scossu>ajs6f: I think it was fedora:resource that was being redefined in a new CND snippet, which was assuming the relaxed node type. Also, I appended a random UUID to the node name.12:52
Although you said I should remove fedora:autoVersioned from the default CND, I put it back because it makes sense to have available for any node to add on the fly as a mix-in. What do you think?12:53
<ajs6f>scossu: So that means the change is only meaningful for your branch, right? Or do I misunderstand.12:54
As for fedora:autoVersioned, is that a mixin from the versioning work done before your branch started, or did you add that yourslef?12:55
<scossu>ajs6f: That I cannot swear on. The main branch is passing the tests, but I didn't check if the properties are being set correctly in that test.12:56
ajs6f: I added fedora:autoVersioned myself, although the test had it added at runtime. It is not referenced anywhere else in the default CND, it's just there for who want to add it.12:58
<ajs6f>scossu: I ask because mikeAtUVa did some work on controlling autoversioning behavior, and I don't want to have these thing conflict.
mikeAtUVa: Did you use a content model/node type to represent autoVersioning?
<gregjansen>fasseg: I was running tests just fine, then I started getting this error for every operation:13:01
java.lang.IllegalStateException: Transaction TransactionImple < ac, BasicAction: 0:ffff7f000101:d4d8:5316370d:9fc4 statu13:02
s: ActionStatus.ABORTED > is not in a valid state to be invoking cache operations on.
<scossu>ajs6f, MikeAtUVa: If the change is on master, I must have merged it because I'm building on the latest master stand.
<gregjansen>I am using FileCacheStore for persistence, but on Ubuntu it should support transactions via file locks.. Perhaps I should set up leveldb again.13:03
<ajs6f>scossu: Okay, let's not worry abot it too much, as long as you are past that set of tests and back onto the ones with which you are really concerned.13:04
<pivotal-bot____>Gregory Jansen added comment: "Some results13:08
TCP cluster of 7 nodes, via LB: ""
1000 102400 1 INGEST 130554 0.7480142 no-tx*
TCP cluster o..." https://www.pivotaltracker.com/story/show/65505154
Gregory Jansen added comment: "Now getting persistent infinispan errors when creating new objects: ""13:10
java.lang.IllegalStateException: Tran..." https://www.pivotaltracker.com/story/show/65505154
Gregory Jansen added comment: "Consolidating SpringContextSingleton into http-commons test jar." https://www.pivotaltracker.com/story/show/6645409613:23
* github-ff joins
[fcrepo4] ajs6f pushed 2 new commits to master: http://git.io/yDL1fw
fcrepo4/master 5c96e1f Kai Sternad: Remove unused code from SessionFactory...
fcrepo4/master 4c51dba ajs6f: Merge pull request #262 from kaisternad/fcrepo-65933082...
* github-ff leaves
<pivotal-bot____>A. "Torcrexilly" Soroka added comment: "Merged with:
https://github.com/futures/fcrepo4/commit/4c51dbaaee311bbdbb5d948945198636c507d268" https://www.pivotaltracker.com/story/show/65933082
A. "Torcrexilly" Soroka delivered "Remove dead SessionFactory code" https://www.pivotaltracker.com/story/show/65933082
<mikeAtUVa>ajs6f, scossu: at one point I recommended in the documentation (and maybe an integration test) that a mixin be created that autocreate a key property that indicates a node should be versioned each update. It's still a reasonable way to enable automatic versioning.13:24
<ajs6f>mikeAtUVa: But there is no sch thing right now?13:25
<mikeAtUVa>ajs6f: You're correct, the default CND in git does not include such a mixin.13:27
<ajs6f>scossu, mikeAtUVa: Okay, it sounds like scossu _could_ leave it in or take it out. It's not connected to anything, but it is something that we thought we should probably put in at some point. But given that it's not really germane to what scossu is trying to do, I would leave it out for now.13:28
* dwilcox_ leaves13:29
<mikeAtUVa>ajs6f, scossu: sounds reasonable... I think it's a reasonable line to draw that unused mixins that are examples should live in documentation not the code base.
<ajs6f>mikeAtUVA: +1
<scossu>ajs6f, MikeAtUVa: Sounds OK.13:34
* travis-ci joins13:40
[travis-ci] futures/fcrepo4#1606 (master - 4c51dba : ajs6f): The build passed.
[travis-ci] Change view : https://github.com/futures/fcrepo4/compare/d887b2154d27...4c51dbaaee31
[travis-ci] Build details : http://travis-ci.org/futures/fcrepo4/builds/20153649
* travis-ci leaves
* dwilcox joins13:41
<pivotal-bot____>Gregory Jansen added comment: "PR updated to move SpringContextSingleton" https://www.pivotaltracker.com/story/show/6645409613:48
<ajs6f>gregjansen: Where did you move it to?13:50
<gregjansen>http-commons org.fcrepo.http.commons.test.util13:51
the projects that need it depend upon the test-jar13:52
<ajs6f>Because it's doing something with HTTP (like setting up a container)?
<gregjansen>yeah, it is usually only needed when the container starts the spring context
so, in a container
<ajs6f>k
Cool.
gregjansen: Is that PR waiting for review?13:54
* cbeer leaves13:56
* gregjansen leaves13:57
* pivotal-bot____ leaves13:58
* pivotal-bot____ joins
* cbeer joins
* pivotal-bot____ leaves14:00
* pivotal-bot____ joins14:01
* hongkong joins14:02
* gregjansen joins
<ajs6f>gregjansen: ping?14:24
* gregjansen leaves14:36
* hongkong leaves14:37
* ajs6f leaves14:45
* ajs6f joins14:47
* gregjansen joins14:55
* gregjansen leaves15:04
* ajs6f leaves15:16
* ajs6f1 joins
* gregjansen joins15:17
ajs6f1: PING!15:20
<ajs6f1>gregjansen: is that PR ^^^ ready for review?
<gregjansen>Yeah
<ajs6f1>Cool.15:22
* scossu leaves15:23
* scossu joins15:36
<pivotal-bot____>Gregory Jansen added comment: "Our largest anticipated ingest next year is a 10TB collection. We'd like to be able to handle such a coll..." https://www.pivotaltracker.com/story/show/6550482415:39
Gregory Jansen added comment: "Individual files in this collection will include HD video files that are several hours in length. So ther..." https://www.pivotaltracker.com/story/show/6550482415:48
<benpennell>I'm not seeing a good way to make fedora prompt for login based on a parameter (as described in https://www.pivotaltracker.com/story/show/65574666 ), since the 403 response is generated by AccessDeniedExceptionMapper, which only has access to an AccessDeniedException thrown by modeshape after AuthorizationProvider.hasPermission() returns false15:56
<pivotal-bot____>feature: Determine authentication challenge strategy for environment with open and protected content. (unstarted) / owner:
* dwilcox leaves15:57
<benpennell>my initial thought was to throw a different exception for a different exception mapper, but AuthorizationProvider.hasPermission isn't expected to throw any exceptions so it'd be outside the api and perhaps not reliable15:59
* ajs6f joins16:02
* ajs6f1 leaves
* ksclarke1 joins16:11
* scossu leaves16:17
* scossu joins16:21
* ksclarke leaves
* ajs6f1 joins16:23
* ajs6f leaves16:24
* ajs6f joins16:38
* ajs6f1 leaves
* kulminator joins16:39
* kulminator leaves16:43
* ksclarke1 leaves16:53
* ksclarke1 joins16:54
* ajs6f1 joins16:55
* ajs6f leaves
* dwilcox joins16:58
* dwilcox_ joins16:59
* ajs6f1 leaves17:02
* dwilcox leaves17:03
* gregjansen leaves
* ajs6f joins17:07
* ajs6f leaves
* dwilcox_ leaves17:08
* dwilcox joins17:09
* mikeAtUVa leaves
* dwilcox leaves17:13
* scossu leaves17:43
* dwilcox joins18:09
* dwilcox leaves18:14
* ksclarke1 leaves20:25
* scossu joins22:45
* scossu leaves22:55