roachtest: decommission test makes improper usage of --join flags

[irfansharif]

Observed in #51649 (review), fall out from #51245 (and I think similar to #51557).

Jira issue: CRDB-4020

[blathers-crl]

Hi @irfansharif, please add a C-ategory label to your issue. Check out the label system docs.

While you're here, please consider adding an A- label to help keep our repository tidy.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is otan.}

[irfansharif]

Hm, this test is written a bit strangely.

cockroach/pkg/cmd/roachtest/decommission.go

Lines 85 to 87 in 4eac05a

	for i := 1; i <= numDecom; i++ {
	c.Start(ctx, t, c.Node(i), startArgs(fmt.Sprintf("-a=--attrs=node%d", i)))
	}

It starts nodes of one-by-one in order to configure specific attributes. Except that doesn't play well with #51329 and #51557, where in trying to avoid a --skip-init flag like was originally attempted in #51329, we auto-init through roachprod if (a) starting a single node cluster (roachprod uses start-single-node, which auto-inits) and (b) we init if starting n1 where n1 isn't already previously initialized.

I can either re-work this test to somehow work around roachprod's auto-init behavior (by starting all nodes and stopping them, and then restarting one-by-one in adding the attributes), but that feels dirty. Alternatively I can have the start-single-node through roachprod similarly persist the cluster-bootstrapped file to prevent re-auto-init? Both of these feel like hacks on hacks that I'd be a bit happier with --skip-init.

[irfansharif]

Oh it's not just that, we're also wiping the entire node away, which wipes out the cluster-bootstrapped on-disk marker.

cockroach/pkg/cmd/roachtest/decommission.go

Lines 244 to 246 in 4eac05a

	if err := c.RunE(ctx, c.Node(node), "rm -rf {store-dir}"); err != nil {
	return err
	}

[github-actions]

We have marked this issue as stale because it has been inactive for
18 months. If this issue is still relevant, removing the stale label
or adding a comment will keep it active. Otherwise, we'll close it in
10 days to keep the issue queue tidy. Thank you for your contribution
to CockroachDB!