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

Using timezone: Eastern Standard Time
(note for the list: if we want to push this planning call earlier, i can live with that. preferrably no more than 2 hours earlier.)11:14
(i'll send that as a discussion item to ff-internal.. some day)
eddies1: if we have a pretty minimal, but deployable databank repository, that's our minimum output. not much development needed, just tweaking integration11:15
eddies1: also, in parallel, deliver fedora 3/glacier. don't let databank stop asynchronous storage efforts.11:16
eddies1: interested in modeshape as a baseline/strawman for scaling performance11:19
eddies1: modeshape itself is too heavy-weight for fedora 4 effort, brings in features no one has articulated as requirements.11:21
<cbeer>eddies1: fixtures for ingest, retrieval, update, etc, with different kinds of content we want to deal with
<barmintor>+1 for following11:25
Eddie asks whether there are any questions
<eddies1>everyone having problems hearing me?
<cbeer>eddies1: yes.
<barmintor>(I'm also getting highly variable volume)
<eddies1>i will reconnect11:26
<cbeer>eddies1: get fedora/akubra working in sprint 1; get databank/akubra working in sprint 111:29
* eddies is an unreasonable product owner.
eddies1: ongoing development of the toolchain outside of fedora (e.g. to support hydra, islandora, escidoc)
<barmintor>is this ActiveRepo gem?11:31
<cbeer>(note: i'm spending the week of MLK working with UCSD, should give the hydra work an additional data point, an escape hatch against databank progress)
barmintor: which may look just like ActiveFedora, with less fedora.
<cbeer>eddies1: stand up hydra, islandora, escidoc against fedora 4.
(barmintor: oh, i remembered my other question from yesterday.. is columbia ok with the FF stuff going onto gluck?)11:32
<barmintor>content, yes- not sure about accounts
I'll email Thomas right now11:33
<cbeer>barmintor: that's fine. we didn't give accounts to many hydra people either
<cbeer>(i've added a link to tracker on the sprint 1 page)11:44
<barmintor>cbeer: just blogging- libxml2, libxslt, mysql, redis-server and solr are all available as homebrew pkgs
<cbeer>TIMEBOX: 1 hour, 15 minutes. your time starts... now!11:46
(copied stories from policy-driven storage to akubra glacier, and deleted policy-driven storage)11:50
epic'd existing sprint and icebox tickets.11:53
+1 to keeping them separate.11:56
* created 6 epics in pivotal tracker11:57
<JasonDGI>connection problems, give me a second12:01
im having trouble reconnecting to the call12:04
<eddies1>jason, what's your skype handle?12:07
i can just bridge you in
<barmintor>eddies1: you're cutting in and out pretty severely12:26
<cbeer>* eddies is still an unreasonable product owner.12:27
* eddies stopped being an unreasonable product owner.12:28
The current release of DataBank (1.0...for marketing reasons) has been held as a last minute bug was encountered. It should be fixed this week and www.dataflow.ox.ac.uk updated with the new download information.12:32
<cbeer>here's the UCSD doc: https://wiki.duraspace.org/download/attachments/33237782/dlpg-DAMS4RESTAPI-131212-1637-20.pdf12:46
<barmintor>did I drop, or did eddies1 drop?12:51
<eddies1>you dropped
* fasseg joins
<barmintor>afk a sec13:00
<JasonDGI>+1 for logic13:03
<cbeer>https://github.com/ucsdlib/dams/wiki/REST+API (replacing the PDF i linked to earlier)13:09
vrbo address: https://maps.google.com/maps?f=q&source=s_q&hl=en&geocode=&q=N+North+Park+Avenue+%26+w.+schiller+st&aq=&sll=41.90885,-87.636775&sspn=0.313758,0.695572&vpsrc=0&ie=UTF8&hq=&hnear=W+Schiller+St+%26+N+North+Park+Avenue,+Chicago,+Cook,+Illinois+60610&t=m&z=1613:14
<barmintor>eddies1: looking at flights arriving the 5th and departing the 14th13:22
<eddies1>i think i just lost my connection13:27
<barmintor>good timing
<eddies1>anyway, thanks everyone and will talk to you tomorrow
<fasseg>see you tommorrow guys
<eddies1>yes, skype gods are hinting
<cbeer>updated the calendar to reflect a 3 hour call.
<cbeer>and sent my email offering an earlier call
<eddies1>incidentally, i'm seriously considering adjusting my sleep schedule for the next six months: bed at 4am and start at noon, that way i can be on for more ppl's waking hours13:29
* eddies1 leaves13:31
* eddies joins13:32
* eddies leaves
* eddies joins
* fasseg leaves13:51
<cbeer>eddies: i'm starting my day at 6am PST now, more or less, for the same reasons.14:52
* JasonDGI leaves15:27
<cbeer>barmintor: i'm thinking about requiring the FF fixtures to be bag-it directories of objects16:18
if only to allow external references for large files
thoughts? better ideas?
<barmintor>cbeer: that's actually how I do Fedora ingest now16:19
<cbeer>cool. then i'll do that
it doesn't seem like a very high bar
<barmintor>I'll say this for macports: it's faster than brew17:04
<cbeer>macports distributes binaries, right?
for better or worse :P
what truly awful graphs.17:08
fedora has such a weird ingest performance profile17:11
it either happens <40ms or more like 500++ms.
<barmintor>cbeer: btw, updating brew => brew installs solr 417:16
<cbeer>great. i should have checked what databank was using before suggesting solr 4
it's just reflex these days :)
<barmintor>what are the chances they're putting unescaped slashes in their field values?17:17
Ahh, but databank has a schema and solrconfig17:20
<cbeer>yeah. not hard to update though17:22
what is it? solr 1.4? 3.x?
<cbeer>ah, then not hard at all. hopefully
want to make a ticket for it? i could try to do it today. or tomorrow morning.
<barmintor>I'll do it right now17:23
<cbeer>that works too. wasn't sure how much longer you wanted to work
so, from this run of modeshape vs fedora:
modeshape: avg 110ms, median 13ms, 370 (add object + binary content cycles?)/minute17:24
fedora: avg 209ms, median 37ms, 283 (create object, add datastream cycles?)/minute
i don't see a way to get percentiles out of jmeter, so i'm going to run it again and save the data somewhere17:25
<barmintor>that's better than I expected from fedora, actually
in relative terms, I mean
<cbeer>it was with a healthy mix of files too. a 30M, 1.3M and 2.8K
<barmintor>modeshape must have some big response peaks, too17:26
<cbeer>max of 346ms
fedora's max was 875ms
and had a lot of requests > 500ms
<barmintor>I wonder how much of that would eventually calm down from GC optimization
for modeshape, anyway
<cbeer>yeah, this is just with jetty defaults, more or less
i'm using hydra-jetty defaults for the fedora tests
and i don't know what tweaks ajs6f may have made to the modeshape jetty
aka, by no means scientific17:28
note to self: when running multiple back-to-back jmeter tests, sometimes it's a good idea to clean out the data folder first17:36
before you run out of disk space.
<barmintor>github lets you branch from the currently viewed context in the web interface now17:37
<cbeer>and delete branches too17:38
<barmintor>pretty nice for doc fixes, need to play with that
* barmintor leaves17:42
<cbeer>hm. i'm dismayed that the only databank tests seem to be full-stack integration tests17:46
all those absolute paths in databank config are going to drive me crazy17:55
this feel so much like rails before rails engines were a thing.17:56
* eddies leaves23:52
* eddies joins23:53
* eddies leaves
* eddies joins

Generated by Sualtam