HDDS-14896. Use separate container for clients in rolling-upgrade suite

[dombizita]

What changes were proposed in this pull request?

Added a separate container for clients in rolling-upgrade suite, so a workaround can be removed, where we checked to not use for tests the SCM container which is currently restarted. To be able to run in an exact container the commands I added that parameter to the helper methods.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-14896

How was this patch tested?

CI on my fork with running the rolling restart test: https://github.com/dombizita/ozone/actions/runs/23915743686

[Russole]

Thanks @dombizita for working on this. The change looks good, LGTM!

[errose28]

Thanks for adding this. Just a few comments, and there are merge conflicts since the last acceptance test change went in.

[errose28]

LGTM, but there's a test failure in the networking. Not sure if it's intermittent.

--- RUNNING ROLLING UPGRADE TEST FROM 2.2.0 TO 2.2.0 ---
Using Docker Compose v2
--- SETTING UP OLD VERSION 2.2.0 ---
--- RUNNING WITH OLD VERSION 2.2.0 ---
Error response from daemon: failed to set up container networking: Address already in use

[errose28]

LGTM, thanks for the updates.

[dombizita]

Thanks for the review @errose28! Just to have it here for later, one issue was that the IP address for the client was the same as one of the multiple S3 gateways after the merge of that. I also saw that I left hostname: scm in the client container yaml and once I removed it it started failing. That hostname is used for the kinit with test user principal, so I added client to init-kdc.sh. After this I realised that if I use the common rolling_restart_service it'll generate data during the stopped period, which is not needed for the client container, so I did a simple stop-start.