Time
S
Nick
Message
07:29 Virgile joined #dataverse
07:56 Virgile joined #dataverse
08:08 Virgile joined #dataverse
08:17 Virgile joined #dataverse
08:33 Virgile joined #dataverse
09:03 Virgile joined #dataverse
09:20 vineet joined #dataverse
10:40 Virgile joined #dataverse
10:43 vineet left #dataverse
10:43 vineet joined #dataverse
10:48 vineet
If this is a personal development installation, we recommend that you undeploy the currently-running copy of the application, with the following asadmin command: /usr/local/payara5/bin/asadmin undeploy dataverse- before attempting to deploy from your development environment in NetBeans.
10:49 vineet
Hello Everyone
10:49 vineet
What does this output means?
10:57 poikilotherm
Hi vineet
10:57 poikilotherm
What are you trying to do?
11:52 vineet left #dataverse
13:15 donsizemore joined #dataverse
13:24 juancorr joined #dataverse
13:26 Virgile joined #dataverse
14:58 pdurbin joined #dataverse
15:48 pdurbin
poikilotherm: good question
15:49 poikilotherm
But they didn't respond :-(
16:07 pdurbin
Yeah. They've been active on the mailing list though. Will probably be back.
17:48 donsizemore
@pdurbin suppose I were throwing together a script to time response deltas from a test dataverse.
17:49 donsizemore
@pdurbin and there are some pretty obvious targets like '/' and '/dataverse/root/?q=test'
17:49 donsizemore
@pdurbin but without being able to predict FAKE DOIs... unless I make some DB calls to grab them?
17:55 donsizemore
@pdurbin or should I just call the contents of the root dataverse and walk my way down?
19:01 pameyer joined #dataverse
19:02 pameyer
@donsizemore what about using the API to get DOIs, and use those for the url construction?
19:02 pdurbin
donsizemore: yes, I'd start by getting the contents.
19:11 donsizemore
I'm gonna pick on one sub-dataverse and get ITS contents
19:12 donsizemore
dvc737e44c to be specific
19:13 donsizemore
what else would you want to gather metrics on? plan was to call each twice, once before cache, once with
19:13 donsizemore
deployment time will need to go in install.py
19:15 pdurbin
Hmm. Well, datasets with lots of versions are slow, I think.
19:15 pameyer
caching where?
19:15 pdurbin
Has Kevin weighed in on this? He might have some ideas for some good perf tests.
19:16 donsizemore
he had some locust stuff i was hoping he'd send me
19:16 donsizemore
@pameyer homepage, primarily.
19:17 pameyer
when I looked at web layer caching, I ran into cookie issues
19:18 pameyer
this may have changed though
19:19 donsizemore
I just mean measuring the difference between the first page load after launch, and the second.
19:19 pdurbin
There's some stuff about Locust here: https://guides.dataverse.org/en/5.3/developers/testing.html#locust
19:21 pdurbin
And a JMeter script from Paul Boon at https://github.com/IQSS/dataverse/issues/6035#issuecomment-514538824
19:22 pdurbin
More on JMeter (a pull request from me, never merged): https://github.com/IQSS/dataverse/pull/6043
19:22 pameyer
@donsizemore same browser session? (still wondering which cache)
19:23 donsizemore
payara cache. just using curl
19:26 pameyer
thanks - I didn't know that payara had a cache
19:29 donsizemore
all I know is the 1st homepage load after launch takes ~7sec on my test box, 2nd call takes ~2sec
19:30 pdurbin
Sometimes people talk about the JVM "warming up". Maybe that's part of it.
19:33 donsizemore
I can watch it in top - launch, CPU goes idle. 1st page load makes Payara pretty busy (along with solr and postgres)
19:33 donsizemore
believe me, I'm very familiar with this sequence ;)
19:34 * pdurbin
believes
19:34 pameyer
fresh solr and postgres sessions too?
19:37 donsizemore
it's usually when i'm biting my nails after a reboot, but payara/glassfish seems to be the difference
19:41 pameyer
makes sense - I'd been thinking about queries hitting the filesystem, and those getting into ram. but that's probably not the slow step
19:43 pdurbin
Now that we're on Payara we could take another look at JCache.
19:44 pdurbin
Performance: Investigate use of JSR 107: JCache - Java Temporary Caching API #2100 - https://github.com/IQSS/dataverse/issues/2100
19:46 pdurbin
I think it's a good move to get measurement in place first though. Thanks for working on this donsizemore.
19:48 pameyer
that reminds me of some discussion of jsf / bean caching that I didn't have the background to follow very well
19:48 pameyer
but that might be a better candidate than jit warmup time
19:59 pdurbin
donsizemore: not sure if you've played this yet, or if it's helpful for what you're doing: https://github.com/IQSS/dataverse/blob/v5.3/scripts/tests/ec2-memory-benchmark/README.md
19:59 donsizemore
I remember it fires in dataverse-ansible. I'm just working on grabbing some quick numbers right now.
20:00 pdurbin
ok
21:47 pdurbin left #dataverse
