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

Using timezone: Eastern Standard Time
* nbanks joins01:26
* nbanks leaves01:31
* kaarefc joins01:41
* nbanks joins01:44
* kaarefc leaves01:59
* kaarefc joins02:36
* github-ff joins04:17
[fcrepo4] nigelgbanks pushed 1 new commit to 49205649: http://git.io/2hnHFQ
fcrepo4/49205649 f6bef32 Nigel Banks: fix for update to modeshape removed Stratgey and NamedHint.
* github-ff leaves
<bljenkins>Project fcrepo4 build #752: UNSTABLE in 17 min: http://ci.projectblacklight.org/jenkins/job/fcrepo4/752/04:34
* ff-github joins08:12
[fcrepo-fixity] fasseg pushed 1 new commit to master: http://git.io/GFm08A
fcrepo-fixity/master 92fd7c8 fasseg: fixed retrieve uris test and added //NOSONOAR comment for ArrayList.class since the interface can not be used
* ff-github leaves
<bljenkins>Yippie, build fixed!08:15
Project fcrepo-fixity build #235: FIXED in 3 min 50 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/235/
* jonathangee joins08:22
<pivotal-bot_>Frank Asseg added "Investigate test failure when building fcrepo4 on windows" https://www.pivotaltracker.com/story/show/5122584708:42
* ff-github joins08:46
[fcrepo-fixity] fasseg pushed 1 new commit to master: http://git.io/9FfFJQ
fcrepo-fixity/master 30a4982 fasseg: added method to check for datastream uris
* ff-github leaves
<fasseg>cbeer: I think tomcat crashed on futures6 since redeployment takes ages and i cant reach http://fcrepo4.fcrepo.org/fcrepo/08:56
hmm do i have permissions there lemme check first08:58
yeah i have some credentials. sorry chris just ignore pls08:59
<bljenkins>Project fcrepo-fixity build #236: FAILURE in 15 min: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/236/09:01
* gregjansen joins09:12
* kaarefc leaves
* jcoyne joins09:25
Chris showed me the latest build of FF. I'm very happy with what ya'll have done.
Good job!09:26
* awoods_ joins09:27
<cbeer>eddies: can you ping me when it's time for standup? i'll try to make the call today09:34
<pivotal-bot_>Unai edited "Investigate test failure when building fcrepo4 on windows" https://www.pivotaltracker.com/story/show/5122584709:47
* ajs6f joins09:50
* jcoyne leaves10:02
* jcoyne joins10:08
<pivotal-bot_>Chris Beer accepted "Get mode-1908 merged for modeshape 3.3" https://www.pivotaltracker.com/story/show/4986809910:15
* barmintor leaves10:24
* ff-github joins10:28
[fcrepo4-demo-hydra-app] cbeer force-pushed master from 8315cef to b03c701: http://git.io/GZvGOQ
fcrepo4-demo-hydra-app/master b03c701 Chris Beer: initial commit
* ff-github leaves
* barmintor joins10:34
<eddies>barmintor: join the party10:45
fasseg & nbanks too10:46
<pivotal-bot_>Unai added comment: "
But when I tried to request it, i get 500 error: ""
http://localhost:8080/rest/objects/0b3925a2-b1a5-437c-ba58-4870..." https://www.pivotaltracker.com/story/show/51225847
<barmintor>I'm standing up! I want the world to know, I got to let it show...
eddies: Sounds like a killer tailgating party10:47
<ajs6f>Jazz hands, barmintor!
<pivotal-bot_>Unai added comment: "After running fcrepo without tests successfully, we create an object: ""10:49
$ ./curl -X POST http://loca..." https://www.pivotaltracker.com/story/show/51225847
Unai edited "Investigate test failure when building fcrepo4 on windows" https://www.pivotaltracker.com/story/show/51225847
* aawoods leaves10:55
<cbeer>fasseg:10:58
there's the MODE API in kitchen sink, right?
why not just use that?
<fasseg>that's the patched modeshape from our futures branch, right? and eddie wanted to get numbers from a vanilla mode installation10:59
<pivotal-bot_>Chris Beer deleted "AF fixes" https://www.pivotaltracker.com/story/show/5110215911:00
<cbeer>fasseg: i don't think any of our changes really have any impact on CRUD operations11:01
<barmintor>eddies: When I get this ticket wrapped up, I'm going to check out these Spring/injection branches and put another set of eyes on it11:07
<cbeer>and now futures6 seems to have disappeared
<ajs6f>barmintor: THANKS!
<pivotal-bot_>Nigel Banks added comment: "https://github.com/nigelgbanks/fcrepo4/tree/49205649-injectable-sessions-and-49012389-workspace" https://www.pivotaltracker.com/story/show/4901238911:10
Nigel Banks added comment: "https://github.com/nigelgbanks/fcrepo4/tree/49205649-injectable-sessions-and-49012389-workspace
" https://www.pivotaltracker.com/story/show/49205649
Andrew Woods edited "Inject authenticated sessions into the JAX-RS resources" https://www.pivotaltracker.com/story/show/4920564911:11
<barmintor>eddies, awoods, et al: If you want to hear a little more about this ISPN stuff, stay on the line and I'll talk it out a bit
* awoods joins11:12
<fasseg>eddies: So you think it's ok to run the tests versus the kitchen sink modeshape instance or should I take the time to get a vanilla instance up and running?
<ajs6f>fasseg/eddies: The real difference will be in MODE config itself.11:13
<cbeer>(ran out of battery on the laptop.11:14
<pivotal-bot_>Nigel Banks started "make islandora work with latest build of fcrepo4. " https://www.pivotaltracker.com/story/show/4835140111:16
<fasseg>tomcat on fcrepo4.fcrepo.org gone again..?11:21
<cbeer>fasseg: futures6 is gone
<fasseg>kk11:22
<cbeer>i've asked anusha to take a look
awoods: no11:27
awoods: on futures2, or your laptop?11:29
eddies: maybe we should try spinning something up on AWS or something
at least take the oxford infrastructure out of the equation11:30
eddies: i just don't trust that oxford's not throttling us or something.11:31
<ajs6f>+1 to taking the metal out of consideration.
<awoods>https://wiki.duraspace.org/display/FF/tmp+results11:36
<eddies>AWS Large, 64-bit, Amazon Linux, Oracle Java 7, Provisioned IOPS at 500?11:39
https://wiki.duraspace.org/display/FF/Fedora+4+test+results+-+2013-06-0311:52
<barmintor>eddies: nope12:00
Have you ever *really* thought about read? I mean, wow. WOW. It's, like, you're totally plugged into the data. Wow.12:01
<ajs6f>Put down the bong, barmintor.12:03
And pass it over here.
<pivotal-bot_>A. "Codiselic" Soroka added "Examine (profile) update data operations." https://www.pivotaltracker.com/story/show/5124224312:05
A. "Codiselic" Soroka started "Examine (profile) update data operations." https://www.pivotaltracker.com/story/show/51242243
A. "Codiselic" Soroka added comment: "Note that the JMeter scripts to run tests and R scripts to analyze them are in two different GitHu..." https://www.pivotaltracker.com/story/show/5124224312:07
<barmintor>get to work, codemonkeys!12:11
<ajs6f>I am not a code monkey. I am a code octopus.12:12
<barmintor>I did that.12:15
afk a sec12:19
* github-ff joins12:26
[ff-jmeter-madness] awoods pushed 1 new commit to master: http://git.io/R3ur3Q
ff-jmeter-madness/master e9ec804 Andrew Woods: Update ReadMe and normalize test plan.
* github-ff leaves
* github-ff joins
[fcrepo4] ajs6f created OltuOAuth (+1 new commit): http://git.io/fuKzrA
fcrepo4/OltuOAuth 6c51414 ajs6f: Further thrashing...
* github-ff leaves
<fasseg>updated https://wiki.duraspace.org/pages/viewpage.action?pageId=3465367312:33
^^ profiling results and procedure
<awoods>fasseg: thanks12:39
* nbanks_ joins12:41
<ajs6f>The README for ff-jmeter-madness directs us to use the kitchen-sink build for testing. Is that right? Should we go against the fcrepo-webapp?
<cbeer>ajs6f: i think it says that so you can use the fcrepo3 legacy api12:42
but i think awoods/fasseg are using the new api
<ajs6f>Do we care about that?
Right.
* jcoyne leaves
<ajs6f>So I'm going to go ahead and use the fcrepo-webapp...12:43
Also, there's a https://github.com/futures/ff-jmeter-testResults/blob/master/fedora-jmx.r in both ff-jmeter-madness and in ff-jmeter-testResults. Anybody know which is the right one to use, or whether it matters?
<awoods>either webapp is probably fine for now
<ajs6f>Well allriiiigghhty then.12:44
* nbanks leaves
<awoods>ajs6f: there are the test scripts (https://github.com/futures/ff-jmeter-madness/blob/master/plans/fedora.jmx) and the R scripts (https://github.com/futures/ff-jmeter-testResults/blob/master/2013-06-03-Fedora4Tests/fedora-jmx.r)12:45
<ajs6f>Right, so what is:
https://github.com/futures/ff-jmeter-madness/blob/master/fedora-jmx.r12:46
?
<awoods>R scripts
<ajs6f>Yes— but which one to use? Or does it matter?
<awoods>but the R scripts need to be tweaked for different tests/outputs
<ajs6f>Which is documented in the script, perhaps? {crosses fingers}12:47
<awoods>the top-level R script can be ignored. the R scripts in the sub-directories should be used with given tests.
what documentation are you looking for, ajs6f?12:48
how to tweak the script?
<ajs6f>Yes.
<awoods>not going to find it
<ajs6f>Right on, Donkey Kong.
Well, I've got to get the tests running and
<awoods>however, to simply run the script:
<ajs6f>profiled first, anyway. 'm jumoing ahead of myself.
<awoods>R --verbose12:49
$ source('fedora-jmx.r')
<ajs6f>Yeah, I'm find with R.
fine
<awoods>then you should be fine tweaking
<ajs6f>You don't think we should document this?12:50
I guess I'll do that after I figure out how to do it.
<awoods>better documentation is always better12:51
* nbanks_ leaves12:57
* nbanks joins12:58
<ajs6f>Okay, I'm missing something. I just ran the JMeter with:13:17
jmeter -n -t plans/fedora.jmx -Jfedora_4_server=localhost -Jfedora_4_context='rest' -Jnum_threads=1 -Jloop_count=3000 -Jfilesize_mean=1MB -Jfilesize_stddev=1
and it ran perfectly well, but it did nothing but create objects. Is there a flag to throw to run the full cycle?
<awoods>no, I do not believe there is a "flag"...13:25
open the jmeter UI: jmeter13:26
<ajs6f>Oh, you have to do it manually, by turning on thread groups?
<awoods>then with ctl-t you can toggle what is on or off
yes
<ajs6f>Someone want to write a ticket for me to parameterize that?
<awoods>sure13:27
<ajs6f>Bless you, sonny.
<awoods>thank you, poppy.
<pivotal-bot_>Andrew Woods added "Parameterize JMeter thread groups" https://www.pivotaltracker.com/story/show/5124857313:28
<ajs6f>Okay, do we have any record of which modules/threadgroups have to be turned on together? I tried turning on what seemed to be an appropriate combo but am just getting a spew of 404s.13:36
<awoods>to run all fedora4 tests (create/read/delete), turn on: "Create Fixtures", "Fedora 4", "Destroy Fixtures"13:43
and make sure "Fedora 4"/"Fedora Module" is set to "Fedora Profiling > Fedora Basic Test"13:44
you may have to clear you data directories under mode.13:47
<ajs6f>Thanks much. This goes in the README.
<awoods>That would be great13:48
* github-ff joins13:49
[ff-jmeter-madness] ajs6f pushed 1 new commit to master: http://git.io/hLQlzg
ff-jmeter-madness/master a8b934a ajs6f: Updating README
* github-ff leaves
<ajs6f>Hm. Intriguingly, that did nothing— like it literally sent no requests at all. I guess it's time to start tracing through the JMeter.13:52
<awoods>This works for me: https://wiki.duraspace.org/download/attachments/21463075/Screenshot+from+2013-06-06+13%3A56%3A46.png13:58
<ajs6f>Checking...14:02
Looks identical. {sigh}14:03
<awoods>From what I am seeing when running the "Create New Objects" test, is that almost all of the time is going to "java.io.DataInputStream.readInt()" from jgroups14:04
<ajs6f>That would point to data-in-transit for clustering, I _think_.
<awoods>https://wiki.duraspace.org/download/attachments/21463075/Screenshot+from+2013-06-06+14%3A04%3A31.png14:05
<ajs6f>Wow. That's _all_ the time.14:06
Did the early Chichago results for MODE use clustering?
That might be the big difference.
<awoods>ajs6f: is your "FEDORA_4_SERVER_CONTEXT" set correctly?
<ajs6f>Is that different from "fedora_4_context"?14:07
(As set on the cmdline?)
<awoods>no, same
<ajs6f>No, looks like its the same.
SO yes, it's set.
And I can get obj creation going fine.14:08
<awoods>what does your log say?
<ajs6f>Nothing.
<awoods>no log?
<barmintor>cbeer: offloading the lowlevelentry creation to ISPN caches makes unpacking ChainCacheStores a bit more difficult
<ajs6f>ajs6f-mbpro-3:ff-jmeter-madness ajs6f$ more log/jmeter-fedora4-1MB-mean-1-stddev-1-threads.csv timeStamp,elapsed,label,responseCode,responseMessage,threadName,dataType,success,bytes,Latency
One line of headers.
ajs6f-mbpro-3:ff-jmeter-madness ajs6f$ more log/jmeter-fedora4-8192-mean-1-stddev-1-threads.csv14:09
timeStamp,elapsed,label,responseCode,responseMessage,threadName,dataType,success,bytes,Latency
awwods: Was that run you just showed on the Oxford gear?14:11
<awoods>no, local14:16
<ajs6f>I wonder what it was doing with all that network time. You were running a single instance? I guess it's just one of those weird outliers.14:17
<awoods>ajs6f: you can see what JMeter is doing by selecting the yellow warning symbol in the top right of JMeter UI.
I am not running in a cluster config14:18
<ajs6f>Oh, that's interesting. It's throwing JavaScript errors from the fixture-creation.
Well, I manually created the "fixtures" directory, which got rid of the error, but it still doesn't do anything (no requests fired).14:20
Hm. 2013/06/06 14:17:59 INFO - jmeter.engine.StandardJMeterEngine: No enabled thread groups found
That's certainly not ture.
true
<awoods>clear the internal (yellow icon) log with the double broom.14:21
what error are you seeing in the status log?
<ajs6f>None.14:22
Now I'm seeing 404s again. Urg.
<awoods>is your "context" correct?
the default is "fcrepo/rest"14:23
<ajs6f>Yeah, I've switched it to just "rest", because I"m running the webapp locally.
via mvn.
Hm. but the server isn't showing the traffic.14:24
I must have some other param wrong.
No, localhost and 8080. That's right.
I wish our log showed the actual URL for a request.14:25
But the whole of the log is:14:26
ajs6f-mbpro-3:ff-jmeter-madness ajs6f$ more log/jmeter-fedora4-8192-mean-819-stddev-1-threads.csv
timeStamp,elapsed,label,responseCode,responseMessage,threadName,dataType,success,bytes,Latency
1370543146288,45,Create Object,404,Not Found,Fedora 4 1-1,text,false,1562,45
Which is weird, because it's still showing only an attempt to create objects, when I definitely have the whole suite enabled.
<awoods>It is set to stop on any errors14:27
<ajs6f>But 404 isn't an error.
Right?
It's just a reponse.
Anyway, I'll turn that off (stop-on-error).14:28
Oh, you were right. That does count as an error.
Now I get slews of 404s and 405s.14:29
<awoods>makes sense
<ajs6f>Not to me, but as long as I know what it does I don't care.
Hm. The URL for object creation is:14:30
/${FEDORA_SERVER_CONTEXT}/objects/${pid}
which doesn't seem to include a host or port...
<awoods>it takes the defaults from "Fedora 4"/"HTTP Request Defaults"14:31
<ajs6f>Ah, cool.
Anyone know how to log the actual URLs being requested by JMeter?14:32
<awoods>you are sure that: curl -X POST http://localhost:8080/rest/objects/<;some-pid> works?14:35
<ajs6f>Just checked, yeah. It's find.
fine
I think I need to find out what the URLs being rquested really are.
<awoods>sounds like it
<ajs6f>Uncle Google?
What can you tell me?14:36
Wow. Seems you can't do it from the GUI. You have to use properties files.14:39
* jcoyne joins14:40
<ajs6f>Hm. We just a flash flood watch for Charlottesville, and TS Andrea is still off the coast of Florida. This should be an interesting hurricane season.
<cbeer>well, that explains that.
<ajs6f>Gnomic, cbeer. Very gnomic.14:41
<cbeer>ajs6f: the oxford vm issue
<ajs6f>Oh?
Anusha's email?14:42
<cbeer>ajs6f: i thought anusha cc'ed ff-internal
i guess they failed-over to a different host
but futures6 didn't make it
<ajs6f>Right, she did cc.
I just didn't realize that's what you were referring to.
Well, it's still better than UVa. I checked out a VM for some work a library science student is doing for me for his practicum. To give him access they asked for his SSN.14:43
And it still took weeks.
Okay, my JMeter run works from the cmdline, but not from the GUI. Oh, well, what the hell...14:45
That's good enough for me.
* ff-github joins14:47
[fcrepo4-demo-hydra-app] cbeer pushed 1 new commit to master: http://git.io/BYmZKg
fcrepo4-demo-hydra-app/master 78a7736 Chris Beer: update gemfile
* ff-github leaves
<ajs6f>Hm. Anyone having difficulty attaching YourKit to a mvn-launched fcrepo webapp/14:48
?
<awoods>I am attaching via tomcat14:51
<ajs6f>Yeah, that's what I'm now going to do.
<awoods>I only had to update my /etc/default/tomcat7
<ajs6f>I just pulled a fresh tomcat.
Quickest road forward.
<awoods>JAVA_OPTS="${JAVA_OPTS} -agentpath:/.../yourkit/tmp/yjp-12.0.5/bin/linux-x86-64/libyjpagent.so"14:52
<ajs6f>Oh, you had to set an agent?
<awoods>yes
probably the same via jetty
<ajs6f>Hmm. I didn't have to.
It was able to attach to a completely stock Tomcat.
<awoods>before I added the agent JAVA_OPT, tomcat was not showing up on the list in YourKit14:53
<ajs6f>Yeah, it did show up for me. (?)
<awoods>whatever works
<ajs6f>awoods: Where was that "find hotspot" tool you used in YourKit?14:55
<awoods>you have to save the session first
<ajs6f>A snapshot?14:56
<awoods>yes
<ajs6f>Cool.
<awoods>send along your hotspot screen shot
<ajs6f>Hm. I wonder if I have to make sure the snapshot captures the moment in which I'm interested: update datastream...14:57
awoods: With an open snapshot, where is that "find hotspot" thingy?14:59
* ff-github joins
[fcrepo-fixity] fasseg pushed 1 new commit to master: http://git.io/uQDEyw
fcrepo-fixity/master d655e7e fasseg: added retrieveDatastreamUris for better testability;added test for method
* ff-github leaves
<awoods>CPU tab
<ajs6f>Hm. I don't see anything like that. Maybe I need to upgrade to YK12? Is that what you are using?15:00
<fasseg>ajs6f: i got that in v11 too15:01
<ajs6f>Is it a button or a mennu selection or something?
<awoods>http://www.yourkit.com/docs/12/help/cpu_intro.jsp
you may have to turn on cpu-profiling first (before the snapshot)
<ajs6f>Ah...
That mayb eit
<awoods>fasseg: are you not seeing the JGroups DataInputStream.readInt() hog?15:03
<ajs6f>awoods; That got it. But the only hotspot it found was the same one you found.15:04
So whether or not fasseg finds it, I did.
<awoods>that one is red-hot
thanks, ajs6f15:05
<bljenkins>Project fcrepo-fixity build #237: NOW UNSTABLE in 6 min 38 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/237/
<fasseg>awoods: yes this seem to be coming from the tcpconnection right?
lemme check my snapshot15:06
<awoods>fasseg: yes, although it is not clear to me why JGroups is getting in the way in a non-clustered config.
<fasseg>awoods: and doesnt this stem from sun.reflect.GeneratedMethodAccessor54.invoke(Object, Object[])?15:08
<awoods>fasseg: actually, I am not seeing the sun.reflect line item on my side.15:09
<ajs6f>Well, not surprisingly, it looks like update datastream takes about the same time as create datastream. Maybe a ms longer.
<fasseg>lemme share my snapshot then...
<pivotal-bot_>Chris Beer accepted "MODE-1952" https://www.pivotaltracker.com/story/show/50923155
<cbeer>bwahahaha15:10
ajs6f: so, when an operation deletes a thing and then recreates it, it takes the same time as creating it?
<ajs6f>In that request thread.15:11
<fasseg>hmm i wonder if this might be off the yourKit agent since i think i ran a cou trace
*cpu trace
<ajs6f>I can believe that, if MODE's binary storage is asynch-deleting the old guy.
<fasseg>my snapshot: https://docs.google.com/file/d/0B5nd_qlYdcqyQXQtZ0xNTmptc1U/edit?usp=sharing
<ajs6f>(VIA binary GC)15:12
DO we want to try to track the entire lifecycle of the actions that follow on a single request? That would be good to know, but I'm not sure how we do that, because they depend on knowing the semantics of the keys for blobs, right?
How else would you know that time the GC thread spends deleting an old "version" is part of the service of the original "modify" request?15:13
<fasseg>but im off now for the night...15:14
see you guys tommorow
<awoods>fasseg: cheers15:15
<ajs6f>Come to think of it, that may confuse the results of measuring delete datastream, as well. If you're just measuring time in the request thread, that's not the whole story.15:18
* jcoyne leaves15:19
<awoods>I am playing with removing the jgroups-fcrepo-tcp.xml config...15:22
<cbeer>hm. i'm confused why the java opts on futures6 don't seem to take15:31
<ajs6f>awoods/cbeer: Check my head on this: would it make sense to do a new JMeter script that just loaded random data into the same datastream over and over and profile it with the intention of getting some kind of idea of how expensive binary GC is compared to directly measured HTTP service is?
(Like an average proportion, sort of.)
<awoods>ajs6f: That could be interesting... however, are we seeing GC as a potentially significant impact on our overall performance?15:33
<ajs6f>awoods: That's kind of it: I don't feel like I have any idea whatsoever of how much GC costs, or how it affects request responsiveness.
It may be nothing. It may be a huge burden.
I don't know.
I don't think it's a likely candidate for a real bottleneck.15:34
<awoods>likewise
<ajs6f>In that case, my "check out datastream update" issue can be closed with "It's pretty much no worse than create datastream". :)15:35
* jcoyne joins15:39
* jcoyne leaves15:45
<pivotal-bot_>Chris Beer finished "Create new, vanilla sufia-based hydra head using the fcrepo4 fork of Rubydora and deploy it to futures6" https://www.pivotaltracker.com/story/show/4988138515:47
Chris Beer delivered "Create new, vanilla sufia-based hydra head using the fcrepo4 fork of Rubydora and deploy it to futures6" https://www.pivotaltracker.com/story/show/49881385
Chris Beer started "Deploy vanilla sufia-based hydra head to futures6/hydra.fcrepo.org" https://www.pivotaltracker.com/story/show/49881575
Chris Beer accepted "Deploy vanilla sufia-based hydra head to futures6/hydra.fcrepo.org" https://www.pivotaltracker.com/story/show/49881575
<ajs6f>Hm. This Thread.run() method sure is using up a lot of cycles. I wonder if we could get rid of it… :)15:50
<awoods>;)15:51
* awoods leaves16:06
* jcoyne joins16:07
* jcoyne leaves
* jcoyne joins16:08
* awoods_ joins16:14
* jcoyne leaves16:18
<aawoods>things are looking better: https://wiki.duraspace.org/download/attachments/21463075/ThreadCountVsElapsedTimebyOperation.png16:38
<ajs6f>You mean without clustering?16:39
<aawoods>indeed
<ajs6f>That's not good. We're selling clustering.
<aawoods>clustering performance is not horrible, as seen by the earlier charts... but cluster does not match up to a single-node fedora3 config.16:40
<ajs6f>Cluster performance per node?
<aawoods>We have not tested a real cluster. We have only tested a single node configured for a cluster. However, single-node fedora4 (non-cluster) performance is now considerably better than single-node fedora316:41
<ajs6f>Well, that's good. I think perhaps we really need to test a cluster. We're planning on bragging about it's performance. :)16:42
<aawoods>true, I believe Ben is heading that up.16:43
<ajs6f>Oh, cool.
In the end, it's not about speed. It's about throughput.
<aawoods>and style
<ajs6f>That's not my dep;t.
<aawoods>What!! you are the style bandit!
<ajs6f>then we're in real trouble.16:44
Arg.
<aawoods>style pirate?
<ajs6f>There is no freeaking way to make this injection stuff work! If one class of things get proper injection, another doesn't!
* jcoyne joins16:54
* jcoyne leaves17:05
<ajs6f>barmintor: Here's one for your list of awesome Java type names: ResourceComponentProviderFactory. Even better: it's an annotation!17:08
* barmintor covers his eyes
<ajs6f>I'm going to use it to annotate my BeanConfigurationPropertiesModel<?>.17:09
* github-ff joins17:11
[fcrepo4] barmintor pushed 1 new commit to distexec: http://git.io/3rvJXw
fcrepo4/distexec 94d079a Benjamin Armintor: Add a subtype for chaining stores, and have cluster executions of transforms unroll them for chaining stores
* github-ff leaves
<barmintor>cbeer: ^^ IT and units pass, want to review a PR?17:13
actually, I shoud squash those commits first, that branch is kind of a mess17:14
* gregjansen leaves17:22
* github-ff joins
[fcrepo4] barmintor deleted distexec at 94d079a: http://git.io/g2KqkQ
* github-ff leaves
* github-ff joins17:23
[fcrepo4] barmintor created distexec (+1 new commit): http://git.io/n0hFzw
fcrepo4/distexec c628fde Benjamin Armintor: use DistributedExecutor libs from ISPN to get FixityResults
* github-ff leaves
<bljenkins>Yippie, build fixed!17:24
Project fcrepo4 build #754: FIXED in 12 min: http://ci.projectblacklight.org/jenkins/job/fcrepo4/754/
Project fcrepo-fixity-corrupter build #57: SUCCESS in 55 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity-corrupter/57/17:25
Project fcrepo-fixity build #238: STILL UNSTABLE in 3 min 1 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/238/17:27
Project fcrepo-fixity-corrupter build #58: SUCCESS in 50 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity-corrupter/58/17:36
<ajs6f>One thing that makes me really happy: Jersey logs nothing.
That's so cool.
<barmintor>I HATE FINAL METHODS17:37
<ajs6f>TESTIFY!17:38
<bljenkins>Project fcrepo-fixity build #239: STILL UNSTABLE in 3 min 8 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/239/
<ajs6f>I hate the open for extension, closed for modification principle.
It doesn't work in a world where we spend far more time doing integration programming then model programming.
than
Well, I'm done with this getting Spring to inject OAuth filters thing. I haven't actually done it, but I'm pretty much done with it.17:43
It's over. I lose. It wins.
<barmintor>it's in that branch?
<ajs6f>Which'n?
Not nbanks'.
It's in a combo:17:44
<barmintor>OltoOAuth
?
<ajs6f>Oh, yeah. It's that plus the repo fcrepo-auth-oauth.
IOW, the module doesn't work without the branch.
The branch just brings a new method to SessionFactory to get a Session in a given Workspace.17:45
I'm putting all the OAuth-related state in a separate workspace.
We don't have to do it that way.
That's just how I happened to do it.
So the upshot was: using SPring Security's OAuth, I couldn't get all the contexts to load properly to the various servlets that Spring Secutiry demanded. (It needs the Spring MVC machinery.) And with Oltu, I couldn't get the different kinds of injection to operate sanely.17:47
I feel like Spring and Jersey and I are in some kind of weird nightmarish version of the fight scenes from West Side Story.17:48
* github-ff joins
[fcrepo4] barmintor pushed 1 new commit to distexec: http://git.io/Psr3ZA
fcrepo4/distexec 7e9e885 Benjamin Armintor: fix a bad mocking strategy in GetClusterConfigurationTest test that was reporting a success for the wrong reasons
* github-ff leaves
<ajs6f>They're dancing around waving prop wooden switchblades and mocking me.
I want to go work on a reasonable, tractable problem, like distributed transactions or creating a Web-scale architecture for provenance.17:50
* github-ff joins17:53
[fcrepo4] barmintor opened pull request #72: Distexec (master...distexec) http://git.io/TD7nMw
* github-ff leaves
<pivotal-bot_>Benjamin Armintor finished "Update LowLevelStorageService to be cluster-aware" https://www.pivotaltracker.com/story/show/47730607
Benjamin Armintor added comment: "https://github.com/futures/fcrepo4/pull/72" https://www.pivotaltracker.com/story/show/47730607
<ajs6f>barmintor: cbeer is going to review that, right?
<barmintor>I hope so
<ajs6f>I just mean, you don't need me to. Cause I'll do it. It's something to do that isn't what I'm doing.17:54
And I don't want to work on my OR poster.
<barmintor>cbeer is the requester, but I have a feeling you will want to rename some classes in that branch17:55
I wanted to get a version that worked up, though, and I'm exhausted
so I'm going to relax with some OAuth/Spring
<ajs6f>I'll do that (if I do) after merging. Actually, I usually have more tolerance for batshit type names than you do, so I doubt I'll have any quibbles.17:56
REEEEELLLLAAAXXXXX!?!?!?!?
(How _would_ you pronounce repeated 'X'es?)
<barmintor>ajs6f: modem noise17:57
<ajs6f>*Is now practicing making modem noise in the back of his throat by gargling whiskey and soda.
There's a P.G. Wodehouse story that revolves around the way that an elderly aunt of the protagonist pronounces the letter X as a beauty exercise. I think it was a Mulliner story.17:59
<bljenkins>Project fcrepo-fixity-corrupter build #59: SUCCESS in 1 min 33 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity-corrupter/59/18:00
<barmintor>ajs6f: which test should I be looking at?18:01
<pivotal-bot_>A. "Codiselic" Soroka added comment: "The upshot of this is: a modify operation on a datastream takes almost no more time than a create ..." https://www.pivotaltracker.com/story/show/5124224318:02
A. "Codiselic" Soroka finished "Examine (profile) update data operations." https://www.pivotaltracker.com/story/show/51242243
<bljenkins>Project fcrepo-fixity build #240: STILL UNSTABLE in 3 min 27 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/240/
<ajs6f>They are the same. I never bothered to finish the second (if you look carefully at it, it is the same as the first) because the first didn't work after I engaged the filters.
barmintor: Does that make sense?18:05
<barmintor>ajs6f: No, but I'm just going to look at the diffs
<ajs6f>Okay. Give me a yell if I can help. And thanks.
<barmintor>this is not the right branch18:11
<ajs6f>Eh?
<barmintor>my diff
sorry, ignore me
<ajs6f>Cool.
To be sure, it's OltuOAuth.18:12
<barmintor>yep
<ajs6f>Cool.
TS Andrea is impressive. We're just getting the outermost effects up here, and the rain is unrelenting. My chickens are not happy.18:13
<barmintor>ajs6f: is it significant if the http-api IT runs?18:28
<ajs6f>It always did for me. I never did anything that would bother them.
<barmintor>let me rephrase: How will I know when I'm done?
:)
<ajs6f>It's the IT in the module that fails like a burp in my face.18:29
The IT
<barmintor>"the module"
<ajs6f>fcrepo-auth-oauth
<barmintor>ok, thanks!
<ajs6f>I went at it with the idea that bringing authN in shouldn't change a thing in the core.
It's just a layer on top.18:30
Or should be.
<barmintor>oh
hmm
<ajs6f>Well, that's what we said on some call or another
<barmintor>sure
just going to mvn install18:31
<ajs6f>And I really don't like the idea of OAuth being an necessary part of the HTTP API. For example, I know that at UVa, we won't want it at all.
(Unless InCommon decides to abandon Shibb and "go OAuth".)
On a mvn install, you should see the tests blow out with errors indicating that injection hasn't taken place in the scanned beans.18:32
* github-ff joins
[fcrepo4] barmintor pushed 1 new commit to OltuOAuth: http://git.io/-ir8RQ
fcrepo4/OltuOAuth 5f84b21 Benjamin Armintor: a spring typo, a doc typo
* github-ff leaves
<ajs6f>:)
I fixed that bad plural and forgot to push it.18:33
* bljenkins leaves18:36
* github-ff joins18:45
[fcrepo4] barmintor pushed 1 new commit to OltuOAuth: http://git.io/UFZAZA
fcrepo4/OltuOAuth 7ac01fe Benjamin Armintor: the typo again
* github-ff leaves
* barmintor leaves
* bljenkins joins18:48
Project fcrepo-fixity-corrupter build #61: SUCCESS in 37 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity-corrupter/61/18:57
* ajs6f leaves18:58
<bljenkins>Project fcrepo-fixity build #242: STILL UNSTABLE in 2 min 17 sec: http://ci.projectblacklight.org/jenkins/job/fcrepo-fixity/242/18:59
* bljenkins leaves19:01
* bljenkins joins19:04
* nbanks leaves19:22
* aawoods leaves19:30
* nbanks joins19:53
* nbanks leaves20:13
* jcoyne joins20:14
* jcoyne leaves20:16
* nbanks joins22:50
* nbanks leaves22:55