Time
S
Nick
Message
07:38 jri joined #dataverse
07:47 jri_ joined #dataverse
09:34 juancorr joined #dataverse
11:24 donsizemore joined #dataverse
11:26 donsizemore
@pameyer I briefly ran KDE on Solaris... then I discovered WindowMaker.
12:38 donsizemore
@poikilotherm tacking on to pete's question, where does your production PG server live? (or really, what's your preferred DB backup strategy, I don't find one at https://dataverse-k8s.readthedocs.io )
12:39 poikilotherm
Our PG is running next to the other pods
12:39 poikilotherm
We have two K8s namespaces, having their own postgres servers
12:39 poikilotherm
They are dumped every hour, which is then backed up to outside of the cluster
12:40 donsizemore
excellent
12:40 poikilotherm
Obviously this is completely up to you.
12:40 poikilotherm
You could as well choose to run a shared postgres
12:40 poikilotherm
Or even a HA cluster of em
12:40 poikilotherm
I shall not declare what you need :-D
12:41 poikilotherm
There are lots of places out there, where DBAs are really picky about who runs a database ;-)
12:43 donsizemore
pick yer poisson
12:43 donsizemore
(speaking of stinky fish)
12:44 poikilotherm
yeah
12:45 poikilotherm
there might be completely different opinions about running a postgres db in production and better not interfere with those religions
12:47 donsizemore
i'm about to do some branch cleaning on Odum's fork but you've frightened me
12:52 poikilotherm
Oh?
12:53 donsizemore
though i declined write access to IQSS so i'm just being paranoid
13:51 pameyer joined #dataverse
14:16 pdurbin
Whoops, 40+ unread message. Good morning!
14:16 pdurbin
donsizemore: I'm glad you got a kick out of my "DataCite is Down" dataset.
14:22 donsizemore
just to be sure @pameyer sees: I briefly ran KDE on Solaris... then I discovered WindowMaker.
14:22 donsizemore
anything based on the old NeXT interface is inherently wonderful
14:23 donsizemore
@pdurbin DataCite is a common source of noise for us. user merely sees "403 Forbidden" but server.log sez "DataCite is down"
14:29 pdurbin
donsizemore: this happens when the user clicks Publish?
14:29 donsizemore
yes
14:30 donsizemore
keep in mind we're still on 4.16
14:30 pdurbin
well, the code hasn't changed in a while, I don't think
14:31 pdurbin
hopefully the user experience will be better with 5.0 but I'm not sure what will happen on Publish when DataCite is down
14:32 donsizemore
weeping and gnashing of teeth
14:33 pdurbin
:)
14:33 pdurbin
we have a single point of failure baked in, I guess
14:35 poikilotherm
Morning pdurbin :-)
14:36 poikilotherm
I'm already at construction site, but you might feel tempted to take a look at https://github.com/IQSS/dataverse/pull/7120
14:38 poikilotherm
I'd be glad for feedback if you folks feel this should be included, like I wrote in #7117
14:39 pdurbin
non-opaque?
14:41 pameyer
I should check out the "DataCite is Down" dataset...
14:43 pameyer
@donsizemore the only solaris / solaris-based systems I've used didn't have x11 ... wasn't adventerous enough to think of installing kde or windowmaker on a file server
14:46 donsizemore
I had a Sparc 4, Sparc 5 and Ultra 1 as desktop workstations at varying times
14:58 pdurbin
Does anybody have any ideas on this "400: Bad Request when uploading file" thread? https://groups.google.com/g/dataverse-community/c/ZEZ5tgEoQrk/m/Tbl8029qDAAJ
15:03 bjonnh_ joined #dataverse
15:04 pameyer
on general principles, my first suspision would be nfs file locking with the temporary files
15:04 pameyer
but that likely isn't an issue if it's size dependent
15:10 pdurbin
Hmm. But the whole issue seems to be size-dependent, right? Smaller files work fine? He seems to be hitting a limit somehow.
15:15 pameyer
yeah
16:09 pdurbin
poikilotherm: instead of uri_no (for non-opaque) can we just call it uri?
16:37 jri joined #dataverse
16:41 poikilotherm
Well I could create another field type for URI .
16:42 poikilotherm
URI is a lot more than just absolute non opaque URIs alias URL
16:42 poikilotherm
And there is no option to create further constraints for a field
16:43 poikilotherm
Our use case was specifically about URLs, so we can create clickable links from them.
16:44 poikilotherm
One could argue that this is also true for URIs, but we thought it would be a better suit for automated processing and more familiar for people
16:44 poikilotherm
But if you folks really would go for true URIs, I can discuss with my colleagues
16:44 poikilotherm
Maybe we can relax our requirements
16:45 poikilotherm
Instead of having to go with a fork...
16:45 poikilotherm
I would change URL , too, but this is a huge one. There are a lot of places where rendering logic and more is all about having a http like URL only
16:46 poikilotherm
Looking at the metadata schemas, there is often a help text about providing "a full url, starting with http"
16:57 donsizemore joined #dataverse
17:00 pdurbin
This is a pretty deep topic that I don't know much about. You might have to educate some of us along the way, maybe in an update to the issue description.
17:12 pameyer
there's a large population of folks who assume that a url means http/https, and that no other protocols exist
17:24 donsizemore
@pameyer there is also a population who assume that a URL includes no protocol
17:42 pdurbin
widespread confusion, mass hysteria
17:48 pameyer
cats and dogs living together in peace an harmony, frogs raining from the sky...
17:50 pameyer
I don't know the full history of what poikilotherm's digging into; but it may be traveling upstream
17:55 pdurbin
yeah, not sure
17:56 pdurbin
I'm supportive of the idea of representing rsync URLs or whatever. rsync URIs.
18:17 pameyer
pdurbin: now we just need to get the rest of the world onboard ;)
18:18 pameyer
I assume most browsers would fall over on anything other than http/https - heard that ftp urls are getting depreciated
18:18 pdurbin
!
18:19 pameyer
browsers depreciating them, I mean. I wouldn't expect ftp to actually go away
18:20 pdurbin
expect an angry letter from cyberduck
18:20 donsizemore
@pdurbin don't look now, but those same three tests that failed on payara-5.2020.3 just failed in develop on 5.2020.2
18:20 donsizemore
DatasetsIT.testAddRoles:1146 expected:<10> but was:<7>", "[ERROR] DatasetsIT.testCreateDataset:139 expected:<10> but was:<2>", "[ERROR] DatasetsIT.testSequentialNumberAsIdentifierGenerationStyle:758", "[INFO] ", "[ERROR] Tests run: 130, Failures: 3, Errors: 0, Skipped: 4"
18:21 jri joined #dataverse
18:21 pdurbin
heaven help us if we try to add 10 roles but can only add 7
18:48 jri joined #dataverse
19:03 poikilotherm
pdurbin should I ping someone about this stuff?
19:03 poikilotherm
Like jgautier, ...?
19:03 poikilotherm
I don't think Tania needs to take a look as this is not directly UI related, but happy to do it if you think that would be a good idea
19:04 pdurbin
poikilotherm: I would suggest editing the description of the issue and adding a bulleted list of terms your are using.
19:04 poikilotherm
Terms = URLs we will most likely see?
19:04 pdurbin
like this:
19:05 pdurbin
- url: your definition of a url
19:05 pdurbin
- uri: your definition of a uri
19:05 pdurbin
- etc.
19:05 poikilotherm
Ah so you want me to paraphraze wikipedia :-D
19:05 pdurbin
yes please
19:06 pdurbin
I think it will help us all understand.
19:06 pdurbin
examples of each term would be excellent as well
19:06 poikilotherm
Hmm ok
19:06 poikilotherm
Will write sth and pester you again
19:06 pdurbin
sounds good
19:22 jri joined #dataverse
19:26 jri_ joined #dataverse
19:27 donsizemore
@pdurbin on #7089 I know you want concrete numbers, but it occurs to me that Ansible could simply install prometheus/grafana before calling integration tests
19:30 pdurbin
donsizemore: good idea. A good excuse to get my hands dirty with those tools.
19:30 pdurbin
Oh, while you're thinking about that issue, how trivial is it to spin up Dataverse 4.20 on Glassfish 4?
19:30 pdurbin
Using dataverse-ansible, I mean.
19:31 pdurbin
Because (from Slack) the idea for that issue is to compare 4.20 to 5.0.
19:36 donsizemore
@pdurbin you just want a separate group_vars. set dataverse_branch to release and zipurl back to glassfish-4.1
19:36 donsizemore
@pdurbin I did that several times in testing for #6961
19:39 donsizemore
zipurl: http://dlc-cdn.sun.com/glassfish/4.1/release/glassfish-4.1.zip zipchecksum: sha256:3edc5fc72b8be241a53eae83c22f274479d70e15bdfba7ba2302da5260f23e9d
19:43 poikilotherm
pdurbin there you go again :-D https://github.com/IQSS/dataverse/issues/7117
19:44 pdurbin
donsizemore: thanks, I was hoping it would be easy like this.
19:45 pdurbin
poikilotherm: typo: vesry
19:46 * poikilotherm
smashed the typo bug
19:47 pdurbin
I can imagine s3:// being a real use case for many.
19:49 pdurbin
I like the mailto:john.doe
19:50 pdurbin
It's weird that jar is supported.
19:52 pdurbin
Is there a word for an HTTP URL ?
19:52 pdurbin
Because you're right, I'm sure all over the code we assume that URLs are HTTP -only.
19:54 pdurbin
I feel like you're saying (and have coded up) that there should be two places for non-opaque URLs: url and uri_no. Is that right?
19:56 poikilotherm
Yes
19:56 poikilotherm
I didn't want to refactor the whole thing and keep the scope narrow...
19:56 poikilotherm
So for me having two implementations felt naturally
19:56 pdurbin
And what goes in one is HTTP URLs. What goes in the other is non-HTTP URLs. Right?
19:56 poikilotherm
Maybe taking a look at https://github.com/IQSS/dataverse/pull/7120/files#diff-d3e306b5540e6ef7d8c2c7f2480fe42a helps
19:57 poikilotherm
Those testcases should be easy to understand :-D
19:57 poikilotherm
you can put any URL in ther other.
19:58 poikilotherm
We want to be able to allow ANY storage location. That might be a HTTP /S based as well
20:01 poikilotherm
Even if you don't want to add ma code, you still could cherry-pick 1e0adac74028aebcac784352f94ffd67c4b7a19c, containing the test refactoring :-D
20:01 poikilotherm
Parameterized test are so much more readable
20:01 pdurbin
Plenty of HTTP URLs in uri_no in those tests. I think conceptually it's easier to tell people, "The url field type now supports non-HTTP urls like rsync, smb, etc." Easier than, "We've introduced a new field type (with a weird name) for non HTTP URLs but don't worry you can put regular HTTP URLs in it too."
20:03 pdurbin
Please think of the poor marketing department. :)
20:06 poikilotherm
OK OK I'll switch LOL
20:06 pdurbin
I mean, get some other opinions.
20:07 pdurbin
That's just my take from a quick look.
20:07 poikilotherm
I did some marketing :-D
20:08 pdurbin
You were in the marketing department? No wonder your Dataverse-on-Kubernetes website looks so good.
20:11 poikilotherm
Just some polishing. Hiding trash under shiny bling bling
20:12 poikilotherm
Getting more contributions in... #17 here I come :-P
20:12 pdurbin
Here's another example for you: x-github-client://openRepo/https://github.com/IQSS/dataverse
20:13 pdurbin
from the "Open with GitHub Desktop" button
20:14 poikilotherm
Oh git:// would be a good thing to add
20:14 pdurbin
yeah
20:16 pdurbin
Can you add a screenshot to the issue of the pain you feel currently when you try to put a rsync (or whatever) URL in Dataverse? The pain is that it isn't a hyperlink?
20:18 poikilotherm
I can create one with the alternative URL metadata field
20:18 poikilotherm
that ok?
20:18 pdurbin
Seems fine. I assume it uses the url field type.
20:20 pameyer
I think git is a better example than rsync - especially since git itself support multiple protocols
20:20 pdurbin
yeah
20:21 poikilotherm
Done
20:23 donsizemore
@pdurbin you're off tomorrow, and i want to rewrite the grafana/prometheus stuff
20:23 pdurbin
yeah, beach day with the (new) dog is the plan
20:30 pameyer
good plan
20:37 pdurbin
See you all on Monday.
20:37 pdurbin left #dataverse
22:16 jri joined #dataverse
22:32 jri joined #dataverse
22:50 jri joined #dataverse
22:52 jri_ joined #dataverse
23:14 dataverse-user joined #dataverse
