4.x: enforce ccm version

[dkropachev]

Recent changes in CCM broke our CICD, let's roll it back to older version.

[Lorak-mmk]

Why not revert / fix ccm instead?

[dkropachev]

Why not revert / fix ccm instead?

It needs to be done, but in CCM repo, what needs to be done here - lock CCM version to one that works

[Lorak-mmk]

lock CCM version to one that works

And then keep it locked because no one remembers to update, and accrue debt, which will at some point make it too difficult to update ever again.

If other projects do this too, then at some point new CCM versions will probably become completely broken too because no one will test them.

[Lorak-mmk]

Reverting broken CCM PR should not be more diffucult than pinning here, and will solve the issue for both this project and others that encounter it.

[dkropachev]

Reverting broken CCM PR should not be more diffucult than pinning here, and will solve the issue for both this project and others that encounter it.

It is not clear what exactly is broken in CCM, by some reason only shard awarness tests are failing, there are 4 commits, none of them looks like one.
But it beside the point, it needs to be pegged to have some stability here.

[dkropachev]

lock CCM version to one that works

And then keep it locked because no one remembers to update, and accrue debt, which will at some point make it too difficult to update ever again.

If other projects do this too, then at some point new CCM versions will probably become completely broken too because no one will test them.

We will update it when we need it, if it works, why bother ?

[Bouncheck]

Seems to work better with CI. I think the consensus around AddedNodeIT was that it's flaky, but I don't know why micrometer test breaks.

I have to agree though that if we move to locked version we will stop noticing breaking changes in master and it will be harder to fix.

[Lorak-mmk]

We will update it when we need it, if it works, why bother ?

We will need to update at some point (for example because older versions may stop working correctly with new Scylla versions). Then it may be much harder to update, because we'll have X months/years of unnoticed errors.
Right now it is enough to revert a broken PR (even if we don't know why it is broken). After a lot of time, we will instead hunt totally unknown errors.