Time
S
Nick
Message
00:10 dataverse-user joined #dataverse
07:34 dataverse-user joined #dataverse
07:34 dataverse-user
.
07:34 dataverse-user
<i>Hey</i>
12:13 donsizemore joined #dataverse
14:28 Virgile joined #dataverse
14:43 Virgile joined #dataverse
18:12 foobarbecue joined #dataverse
18:18 foobarbecue
Hi all & poikilotherm, thanks to your help I have a dataverse up at JPL for testing & demo purposes using k8s, but it's a high priority for us to move to Dataverse 5 for various reasons. Can you give me a feel for how close 5.x support is, and if there's anything I can do to help push it along?
18:29 poikilotherm
My plan was to create images for Dataverse 5 when some stuff in upstream is solved. I want to move the image generation to an upstream repository (the Dataverse codebase), using docker-maven-plugin to create JIB like images. There is already a feature branch for this. But before this is really ready for production use, I want some more stuff done to make the images smaller and less complicated. I don't have a list
18:29 poikilotherm
handy, but maybe I can create one (would be a good idea anyway). I assume this might be a better start for a discussion on how far away we are and what could be left out to speed up the process and where others could chime in to help.
18:30 poikilotherm
I also would like to get rid of the bash scripting glue necessary for configuration etc. I started with https://github.com/GlobalDataverseCommunityConsortium/dvopa and have some ideas for it.
18:31 poikilotherm
This is a very early stage. More people helping to get it going would be great
18:31 poikilotherm
(Technology stack could still be changed)
18:34 poikilotherm
Actually creating images for 5 is not that far away if we cut the development and live a little longer with the situation of scripted glue and complicated inner works.
18:34 poikilotherm
I had hopped to avoid that, but I understand that you need to get things done.
18:35 poikilotherm
Don't tell anyone, but I'm already using the feature branch to create images during dev... ;-)
18:36 poikilotherm
https://github.com/IQSS/dataverse/pull/7245
18:37 poikilotherm
It would be very helpful to think about how we do QA for this. If we want this to live in IQSS/dataverse, @kcondon isn't going to do the QA for this for us. Too much workload on things they don't use.
18:38 poikilotherm
So this could live within the scope of GDCC, but then again: we need to think about how to maintain all of this.
18:38 poikilotherm
There are ideas in my head, but I could use a hand :-D
18:39 poikilotherm
(And maybe someone telling me to cut a release and leave the rest for the next iteration...)
18:41 poikilotherm
Not sure this sheds enough light on all of this for you. We could try to do a VC tomorrow. I could extend my workday and offer a timeslot around 17:15 UTC . No idea what that is in your place.
18:46 foobarbecue
Ok, thanks for the info. We are still at the "gaining traction" stage for dataverse internally -- it's like 25% of my time and even less for 2 other devs. Maybe I'll keep our current setup on dv4 and use it to try and get enough people-hours involved to be helpful to you guys.
18:49 foobarbecue
VC is a great idea, let me loop in the other 2 people involved and find a time soon (does a doodle.com poll work for you)? We're all in CA, PST (10:50am now). What timezone are you?
18:57 poikilotherm
It's CET for me.
18:58 poikilotherm
As I told people here and elsewhere: timezones suck :-/
19:01 foobarbecue
yes... I think we should set up lots of mirrors in space and then we can just switch the planet to simultaneous daylight and all operate on UTC
19:06 poikilotherm
You're actually in a better place and position for that than me :-D
19:06 poikilotherm
I just know a few people at DLR
19:06 poikilotherm
(german aerospace research center)
19:07 foobarbecue
DLR does awesome work. I'm a big fan of their IceMole
19:09 poikilotherm
LOL I just looked at IceMole in WIkipedia an learned that it has been developed mostely by FH Aachen, my "alma mater".
19:12 foobarbecue
Nice. My previous job here was developing an ice climbing robot (IceWorm) which used ice screws and a few months into the project I realized "woah, we're not the first people to put ice screws on a robot"
19:12 poikilotherm
(Ant it's pretty near to where I am and live)
19:12 poikilotherm
Looking from US West Coast it's pretty much the same place :-D
19:13 poikilotherm
It's definitely some crazy tech :-)
19:13 poikilotherm
So much cool stuff going on
19:35 foobarbecue
Indeed. So I have a little glitch on my dataverse instance. If I go to https://mydataverse.com/loginpage.xhtml , all is well. But if I got to https://mydataverse.com/ then it redirects me to https://localhost:8080/loginpage.xhtml?redirectPage=%2Fdataverse.xhtml . Why does it want to send me to localhost? I did set dataverse_siteUrl and
19:35 foobarbecue
dataverse_fqdn in my configmap...
19:36 foobarbecue
(mydataverse.com just represents my domain, of course)
19:44 donsizemore
@foobarbecue what's the siteUrl setting in domain.xml?
19:44 donsizemore
(the installer hard-codes :8080...)
19:45 donsizemore
and which auth method is this?
19:46 foobarbecue
<jvm-options>-Ddataverse.siteUrl=https://dataverse2.jpl.nasa.gov </jvm-options>
19:47 foobarbecue
maybe it needs a trailing slash?
19:47 foobarbecue
local-only auth (default)
19:48 foobarbecue
(note that this domain is internal-only)
19:51 donsizemore
you shouldn't need the trailing slash. is dataverse behind apache, or nginx?
19:51 foobarbecue
apache
19:52 donsizemore
and you've looked at the proxy bits
19:52 foobarbecue
I'm using k3d, with apache forwarding to the k3d built-in load balancer
19:53 donsizemore
and you've made sure there's only one siteUrl setting
19:54 foobarbecue
yep. I also checked `asadmin list-jvm-options` after exec'ing into the running dataverse pod
19:54 donsizemore
anything interesting in the apache logs?
19:54 foobarbecue
checking
19:54 donsizemore
(and you've restarted dataverse after updating siteUrl?)
19:56 donsizemore
in a "plain" installation this should all hinge on siteUrl, unless you were using external auth
19:57 foobarbecue
btw I'm using apache proxypass http://localhost:8080/ instead of using AJP... maybe using AJP would fix the issue? I'll also try a dataverse restart
19:58 foobarbecue
oo I wonder if I need `ProxyPreserveHost On`
19:58 donsizemore
I think you only need AJP for Shibboleth
20:00 poikilotherm
Foobarbecue are you seeing the same problem when using kubectl port-forward to the service directly, circumventing the proxy setup?
20:00 donsizemore
if you keep the proxy, do you have a ProxyPassReverse ?
20:09 foobarbecue
FYI turning ProxyPreserveHost on broke it entirely. I don't have a ProxyPassReverse set. Looking at logs now
20:11 donsizemore
you may simply want a ProxyPassReverse with a value matching ProxyPass
20:21 foobarbecue
donsizemore great thanks, that fixed it!
20:21 donsizemore
excellent!
21:37 sivoais joined #dataverse
21:53 bjonnh joined #dataverse
21:53 sivoais joined #dataverse
21:53 yoh joined #dataverse
21:53 Fishmongering joined #dataverse
21:53 JonathanNeal_ joined #dataverse
21:54 yoh joined #dataverse
21:55 Fishmongering joined #dataverse
22:07 poikilotherm joined #dataverse
