dmwm/WMCore

agent deployment - improve scripts

Closed this issue · 2 comments

mapellidario commented

Impact of the new feature

I would suggest a couple of improvements for the WMAgent deployment procedure:

  • cronjobs should be installed as soon as initializing the agents
  • we should enforce that we do not initialize an agent if the local couchdb is not empty

Is your feature request related to a problem? Please describe.

When deploying a new wmagent, in order to install the cronjobs inside the wmagent docker container, we need to run the script wmagent-docker-run.sh twice, which makes little sense.

Last time Andrea and I deployed an agent, we forgot to wipe the local couchdb. Since we never want to start an agent with a non-empty local couch, we should enforce this check

Describe the solution you'd like

I will open the proper PR to make these changes

Describe alternatives you've considered

We could live as is, but it would be sub-optimal

amaltaro commented

@mapellidario haven't we resolved this issue with the following CMSKubernetes developments:
dmwm/CMSKubernetes#1561
dmwm/CMSKubernetes#1563
and
dmwm/CMSKubernetes#1562
?

If so, then I think this issue can be closed out. In the future, please make a reference from the CMSKubernetes PR to the WMCore GH issue, this way we can easily keep track of these developments. Thanks

mapellidario commented

Ooooops yes we did, thanks alan for the bookkeeping :)

  • 👍1