Page MenuHomePhabricator

Setup "official labs grafana" instance
Closed, ResolvedPublic

Description

Currently the grafana.wikimedia.org installation has no counterpart in the labs infrastructure realm. However, labs does have a working graphite installation on labmon1001.eqiad.wmnet. The installation is used by frontends like nagf (https://tools.wmflabs.org/nagf/) which no matter how nice it is, is not providing the same functionality to labs projects as a dedicated grafana installation would. At the same time, labmon1001 has been added as a data source to the production installation, providing one more argument for the need of a labs grafana install, while at the same time causing some confusion as to how labs projects get to have dashboard at the production installation.

Actionables that IMHO should be implemented:

Get a working grafana installation on labs with a data source pointing to labmon1001
Provide a number of generic dashboards for the various labs projects
Move over to that installation the labs related dashboard that have been created on the production grafana installation
Remove labmon1001 as a data source from the production grafana installation.

11:01 PM <yuvipanda> addshore: there's grafana.wmflabs.org but I've no idea where it's supposed to be running
11:01 PM <addshore> its in the Graphite project
11:01 PM <addshore> grafana.graphite.eqiad.wmflabs
11:02 PM <addshore> It would be cool to get https://github.com/addshore/grafana-wmfpageviews-datasource on it
11:02 PM <yuvipanda> oh
11:02 PM <yuvipanda> addshore: can you file a bug for 'official labs grafana' and I'll set one up from puppet?
11:03 PM <addshore> yes :)
11:04 PM <yuvipanda> addshore: thanks :)

Related Objects

Event Timeline

Addshore assigned this task to yuvipanda.
Addshore raised the priority of this task from to Needs Triage.
Addshore updated the task description. (Show Details)
Addshore added projects: Grafana, Cloud-Services.
Addshore subscribed.
chasemp triaged this task as Medium priority.Dec 7 2015, 8:12 PM
chasemp subscribed.
yuvipanda subscribed.

Unfortunately I don't think I'll have time to do this anytime soon :(

So I just found out that you can use stats from graphite.wmflabs.org on grafana.wmflabs.org - you can switch 'data sources' with the small icon at the bottom right of each graph config. So am using that now.

The answer is 'it depends' I think since edit rights on the grafana.wikimedia.org instance are restricted to a particular LDAP group.

Also my primary case for a labs instance would be to test out new datasource plugins!
Which I guess we are not okay doing on the production version.

I welcome the idea of a Grafana instance dedicated to labs, on the other hand I would like to keep the labs datasource on the production instance. The rational is CI is split between production machines and labs instances to run the tests.

A concrete example is easily switching between the production service Zuul board and the integation labs project overview. That is a quite unique use case though, so I am not going to obsess over it ;-}

Change 299999 had a related patch set uploaded (by Yuvipanda):
grafana: Make role explicitly reference production secrets

https://gerrit.wikimedia.org/r/299999

Change 300004 had a related patch set uploaded (by Yuvipanda):
grafana: Mark role explicitly as production

https://gerrit.wikimedia.org/r/300004

Change 300005 had a related patch set uploaded (by Yuvipanda):
grafana: Refactor production role into base role

https://gerrit.wikimedia.org/r/300005

Change 300007 had a related patch set uploaded (by Yuvipanda):
Add new secrets for grafana labs / prod instances

https://gerrit.wikimedia.org/r/300007

Change 300007 merged by Yuvipanda:
Add new secrets for grafana labs / prod instances

https://gerrit.wikimedia.org/r/300007

Change 299999 merged by Yuvipanda:
grafana: Make role explicitly reference production secrets

https://gerrit.wikimedia.org/r/299999

Change 300004 merged by Yuvipanda:
grafana: Mark role explicitly as production

https://gerrit.wikimedia.org/r/300004

Change 300005 merged by Yuvipanda:
grafana: Refactor production role into base role

https://gerrit.wikimedia.org/r/300005

Change 300020 had a related patch set uploaded (by Yuvipanda):
grafana: Add and provision labs grafana role

https://gerrit.wikimedia.org/r/300020

Change 300021 had a related patch set uploaded (by Yuvipanda):
cache: Add labs grafana behind misc varnish

https://gerrit.wikimedia.org/r/300021

Change 300023 had a related patch set uploaded (by Yuvipanda):
Add grafana-labs and grafana-labs-admin domains

https://gerrit.wikimedia.org/r/300023

Change 300020 merged by Yuvipanda:
grafana: Add and provision labs grafana role

https://gerrit.wikimedia.org/r/300020

This is very close to being done now!

Change 300023 merged by Yuvipanda:
Add grafana-labs and grafana-labs-admin domains

https://gerrit.wikimedia.org/r/300023

Change 300021 merged by Yuvipanda:
cache: Add labs grafana behind misc varnish

https://gerrit.wikimedia.org/r/300021

Change 300076 had a related patch set uploaded (by Yuvipanda):
grafana: Expand edit access in labs grafana

https://gerrit.wikimedia.org/r/300076

Change 300076 merged by Yuvipanda:
grafana: Expand edit access in labs grafana

https://gerrit.wikimedia.org/r/300076

Yesterday spring most probably has broken the labs Shinken that got a 401 trying to reach labmon1001: T140976

Is setup to allow login to https://grafana-labs-admin.wikimedia.org to anyone who is a member of any labs project. Still can't see the dashboards I save there in https;//grafana-labs.wikimedia.org though.

I welcome the idea of a Grafana instance dedicated to labs, on the other hand I would like to keep the labs datasource on the production instance.

Yes this makes total sense, and is already possible.

pasted_file (210×401 px, 17 KB)

Is setup to allow login to https://grafana-labs-admin.wikimedia.org to anyone who is a member of any labs project. Still can't see the dashboards I save there in https;//grafana-labs.wikimedia.org though.

This seems very odd...

@hashar the labs grafana also has prod graphite as a data source...

@hashar the labs grafana also has prod graphite as a data source...

Yup I am well aware about that and I am making use of it. I guess from one of the comment above, I understood the labs source would be dropped. If it is kept available then I have no concern to raise anymore :-}

@hashar no, I meant the opposite - that the grafana at grafana-labs.wikimedia.org also has access to prod graphite - so you can move dashboards that are using labs graphite + prod graphite from prod grafana to labs grafana. I don't know if we'll keep labs graphite on prod grafana forever - @fgiunchedi might have opinions on that.

@hashar no, I meant the opposite - that the grafana at grafana-labs.wikimedia.org also has access to prod graphite - so you can move dashboards that are using labs graphite + prod graphite from prod grafana to labs grafana. I don't know if we'll keep labs graphite on prod grafana forever - @fgiunchedi might have opinions on that.

indeed, once the dashboards graphite-labs as datasource have been moved to grafana-labs I think we should delete the datasource to avoid confusion

@hashar no, I meant the opposite - that the grafana at grafana-labs.wikimedia.org also has access to prod graphite - so you can move dashboards that are using labs graphite + prod graphite from prod grafana to labs grafana. I don't know if we'll keep labs graphite on prod grafana forever - @fgiunchedi might have opinions on that.

indeed, once the dashboards graphite-labs as datasource have been moved to grafana-labs I think we should delete the datasource to avoid confusion

And I second that approach.

I'm going to count this as done, and will open a separate ticket for issues.