arvados/development

Development discussion & coordination for the Arvados project. For Q&A, support and general discussion see https://gitter.im/arvados/community

Lucas Di Pentima

@ldipenti

maybe we could just show the uuid

Yep, things like this:

    <% if obj.is_trashed %>
      <td>
        <%= link_to_if_arvados_object @owners[obj.owner_uuid], friendly_name: true %>
      </td>

Peter Amstutz

@tetron

is it doing something like requesting a specific set of uuids? maybe limit and offset don't need to be there at all

I seem to recall fixing a similar issue

the limit and offset didn't do anything but they still got in the way

Lucas Di Pentima

@ldipenti

It makes sense to query for remote users on a federation, so maybe I should push for omitting those params. The limit sets to the max value possible, and offset sets to 0 if not used

right

Peter Amstutz

@tetron

yes you should omit them

pretty sure there's a way to suppress them from being added by default

I don't remember exactly how

Lucas Di Pentima

@ldipenti

Ok, I’ll keep looking

thanks for the pointer!

Ward Vandewege

@cure

@tetron https://dev.arvados.org/issues/16886#note-14 I'm not sure what to make of this.

Peter Amstutz

@tetron

@cure workbench is behaving strangely too

Ward Vandewege

@cure

how so?

Peter Amstutz

@tetron

if I just leave the dashboard alone, it refreshes and everything is empty. then it refreshes and again and everything is back

Ward Vandewege

@cure

websockets in trouble perhaps?

Peter Amstutz

@tetron

well, if it was flapping between some activate/inactive state that could also account for the error you just noted

Ward Vandewege

@cure

yes!

Peter Amstutz

@tetron

and I did just change something

obviously this wasn't the intended outcome ;-)

Ward Vandewege

@cure

I figured :)

Peter Amstutz

@tetron

I'm going to get lunch then I will dig into it more

Ward Vandewege

@cure

I will keep testing stuff

though... with a bug like this in the mix, that may be futile. Maybe I'll just see how wb2 is doing

Ward Vandewege

@cure

@tetron added a note at the bottom of https://dev.arvados.org/issues/16886#note-14 - seeing the problem in wb2 as well, and it's some sort of permission flapping issue; things work then stop working.

Ward Vandewege

@cure

I'll hold off on more manual testing

Daniel Kutyła

@DannyNs

@tetron https://dev.arvados.org/issues/16812?next_issue_id=16809&prev_issue_id=16816#note-10 tests are ok now thank you @cure @ldipenti for help

Ward Vandewege

@cure

anytime!

Peter Amstutz

@tetron

@cure this issue with is_active flapping between true and false is so weird

Peter Amstutz

@tetron

you can see users getting flipped from is_active: true to is_active: false in the audit log

Lucas Di Pentima

@ldipenti

Could it be related to 16914?

Peter Amstutz

@tetron

@ldipenti yes, that's my theory, I just can't understand why it is happening

the sequence of events seems like, it revalidates the token, it refreshes the user record, and when doing that, it randomly decides to sometimes deactivate the user

Peter Amstutz

@tetron

alright, I got it

easy fix, adding a test

Lucas Di Pentima

@ldipenti

\o/

Peter Amstutz

@tetron

@ldipenti https://dev.arvados.org/issues/16914#note-9 do you want to review the fix

Lucas Di Pentima

@ldipenti

Yes! will do

Agh, go default values biting us again?

Peter Amstutz

@tetron

not exactly

well, it is a problem we have with using structs

but it was a behavior that was already handled for batch updates, I just overlooked it

if you get a user record, but it doesn't have all the records (because you used 'select' in the query) you can't update those fields you didn't get

Lucas Di Pentima

@ldipenti

@tetron 16914 looks great, please merge

Peter Amstutz

@tetron

@cure ok, so I think I fixed the user active bug last night

so you can continue testing

also the build pipeline has a couple of failures so we should probably figure out what is going on https://ci.arvados.org/view/Arvados%20build%20pipeline/

Where communities thrive

People

Repo info

Activity