[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cbi-dev] Interrogations on the new Jenkins/Kubernetes instances

Thanks Denis,

 

Reading you answers I found out why this timed out: I inputed the old Jenkins instance (shared instead of xwt) so now it does what it should do (failing as it does not have write permission it seems but it does seem to work as intended). I will need to open a bug to ask that the write permission for the agent/Jenkins.

So apart from the previous questions, all is normal again in the realm !

 

Cheers,

Quentin

 

De : cbi-dev-bounces@xxxxxxxxxxx <cbi-dev-bounces@xxxxxxxxxxx> De la part de Denis Roy
Envoyà: mardi 11 juin 2019 17:01
à: cbi-dev@xxxxxxxxxxx
Objet : Re: [cbi-dev] Interrogations on the new Jenkins/Kubernetes instances

 

I'm not a release engineer but I can repeat what I've read elsewhere:

 

On 2019-06-11 8:28 a.m., LE MENEZ Quentin wrote:

Hi MikaÃl,

 

I tried correcting the points mentioned but found other problems. Again, if you could help me with them itâd be a great help:

1-      The first point I made was mainly about keeping the workspace for the downloaded maven dependencies which can be a pain to download every single time.(which, if I understood correctly will be the case if the workspace is deleted every single time)

This is an issue and I believe we have ideas for resolving this, but for now, the world needs to be re-downloaded every single time.

 

2-      I tried to add a promote step to the job but I cannot figure out how one would access the archive files from the job as the promote step is launched in another agent. I gather you can get them through a (curl ?) request but that can be a big overhead for larger project archives (e.g. RCPs, zipped p2s containing hundreds of plugins, â). There must be a way to do it cleanly but I didnât find how.

If you're using a pipeline build, you can use stash for this:

https://jenkins.io/doc/pipeline/steps/workflow-basic-steps/

 

 

3-      I tried to publish the contents in the same workspace state (after the build) but the ssh connection did not seem to get through (no error shown just a timeout).

Is the destination still /home/data/httpd/download.eclipse.org or should I call another destination ?

I think, so, based on these docs:

https://wiki.eclipse.org/Jenkins#How_do_I_deploy_artifacts_to_download.eclipse.org.3F

 

I don't have any answers for the rest, but I'm sure Mikael will chime in tomorrow.

 

Denis

 

 

 

As a bonus question it should be possible to use variables in the request but if I transform âssh genie.shared@xxxxxxxxxxxxxxxxxxxxxxxxxxxx mkdir âp /home/data/***â into âssh $sshGenie@$sshRemote mkdir -p $destinationUpdateSiteâ it times out. Is it because the previously mentioned destination is erroneous ?

 

Thanks again,

Quentin

 

Ps : The zip command seem to only be available in the âmigrationâ profile. Should I open a bug for it to be integrated inside the âui-testâ (as I get the feeling that migration is not supposed to last) along the one for Gerrit problems ?

Pps: to provide further information I included the script I intended to use as a promotion script. If you see any kind of problems with it feel free to tell me J

 

 

De : cbi-dev-bounces@xxxxxxxxxxx <cbi-dev-bounces@xxxxxxxxxxx> De la part de MikaÃl Barbero
Envoyà: vendredi 7 juin 2019 09:58
à: Common-build Developers discussion <cbi-dev@xxxxxxxxxxx>
Objet : Re: [cbi-dev] Interrogations on the new Jenkins/Kubernetes instances

 

Hi Quentin,

 

See my answers inline

 




Le 6 juin 2019 Ã 17:58, LE MENEZ Quentin <Quentin.LEMENEZ@xxxxxx> a Ãcrit :

 

Hi Everyone,

 

I might be asking something that is already documented, and if so please direct me to the relevant documentation, but if not here are my questions:

For context I was configuring the XWT jobs on https://ci.eclipse.org/xwt/ and I am not admin of this instance.

 

1-      The xwt-master workspace just vanished without any actions from me (so I gather the message The agent this project has run on for the last time was removed indicates me that the pod/node was used by another job hence the workspace was deleted ?). How so ?

 

Agents on the new infra are spawn on-demand, a new workspace is created at the same time and everything is teared down once the build is done. If you want some files to outlive the duration of the jobs, you should either use the archiveArtifacts Jenkins feature and/or cp the files to the download area (https://wiki.eclipse.org/Jenkins#How_do_I_deploy_artifacts_to_download.eclipse.org.3F). Note that if you push files that are not releases, you should put them in a folder strcuture that will not be mirrored (see https://wiki.eclipse.org/IT_Infrastructure_Doc#Use_mirror_sites.2Fsee_which_mirrors_are_mirroring_my_files.3F for the list of file patterns that are not send to mirrors).




2-      I cannot seem to be able to run the zip command from a shell script build step. Should I use something else ?

 

"zip" command is not available in the default agent image, "unzip" is though. Advanced images are available in "migration" and "ui-test" agents. See https://wiki.eclipse.org/CBI/Jenkins_Migration_FAQ#How_do_I_run_UI-tests_on_the_new_infra.3F for how to use them.




3-      The gerrit Job canât seem to add the +1 to the gerrit contribution even after adding the review parameters 1 and +1. I gather that the Gerrit Reporting Values and Verified Commands are not set in this instance and therefore nothing is sent by the job to the gerrit instance & patch ? 

 

It should work out of the box. Please, open a ticket and we will investigate.




4-      This instance does not seem to have any executors pre-locked. Is/Will this be the future default configuration ? i.e. will they be dispatched by the kubernetes manager and assigned whenever a free node is available ?

 

As stated above, agents / executors are dynamically provisioned. You won't see any until you start a build. Up to 2 agents can be provisioned simultaneously for your Jenkins instance, i.e. you can run up to 2 jobs concurrently.

 

Cheers,

 

MikaÃl Barbero 

Team Lead - Release Engineering | Eclipse Foundation

ð (+33) 642 028 039 | ð @mikbarbero

Eclipse Foundation: The Platform for Open Innovation and Collaboration

 



_______________________________________________
cbi-dev mailing list
cbi-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/cbi-dev

--

Denis Roy

Director, IT Services | Eclipse Foundation, Inc.

Eclipse Foundation: The Platform for Open Innovation and Collaboration

Twitter: @droy_eclipse