roachtest: fix backup2TB by using IMPORT, rip out store dumps #36832
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewed 2 of 2 files at r1, 1 of 2 files at r2, 2 of 2 files at r3.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @petermattis and @tbg)
pkg/cmd/roachtest/backup.go, line 39 at r3 (raw file):
// bank.bank doesn't exist, probably due to some gossip propagation // delay. delay := 5 * time.Second
Can you change this to poll for the existence of bank.bank before launching the BACKUP command? Then a longer timeout, maybe 1 minute, could be used before bailing out.
pkg/cmd/roachtest/cluster.go, line 530 at r3 (raw file):
} func (s *clusterSpec) String() string {
Is this change intentional?
There was a problem hiding this comment.
Reviewable status:
pkg/cmd/roachtest/backup.go, line 39 at r3 (raw file):
Previously, bobvawter (Bob Vawter) wrote…
Can you change this to poll for the existence of
bank.bankbefore launching theBACKUPcommand? Then a longer timeout, maybe 1 minute, could be used before bailing out.
BTW, filed this issue as #36841. This is a bug.
I'll make that change, though I think it's largely an idle diversion. If it takes more than 5s, I'm happy to fail the test. And the test takes many hours, so the 5s don't matter.
pkg/cmd/roachtest/cluster.go, line 530 at r3 (raw file):
Previously, bobvawter (Bob Vawter) wrote…
Is this change intentional?
Yes, with the pointer receiver you couldn't call this when you had a value. Now that it has a value receiver it works both on a clusterSpec and a *clusterSpec. For example, the new code in registerBackup makes use of this.
There was a problem hiding this comment.
Reviewed 2 of 2 files at r1, 2 of 2 files at r2.
Reviewable status:
pkg/cmd/roachtest/backup.go, line 39 at r3 (raw file):
Previously, tbg (Tobias Grieger) wrote…
BTW, filed this issue as #36841. This is a bug.
I'll make that change, though I think it's largely an idle diversion. If it takes more than 5s, I'm happy to fail the test. And the test takes many hours, so the 5s don't matter.
Nit: const delay = 5 * time.Second.
Now that 19.1 is tagged, the store dumps don't work any more. Consequently the tests haven't run in over two weeks. Before that, it had episodes during which it didn't run most of the time due to flakes downloading the store dumps. Hopefully now it'll be more reliable. Fixes cockroachdb#34702. Release note: None
We're unlikely to want to bring them back. Release note: None
It doesn't seem like we're planning on using store dumps in the future. Our efforts should be going into improving IMPORT instead. Release note: None
There was a problem hiding this comment.
TFTR!
bors r=petermattis
Reviewable status:
pkg/cmd/roachtest/backup.go, line 39 at r3 (raw file):
Previously, petermattis (Peter Mattis) wrote…
Nit:
const delay = 5 * time.Second.
I did it The Bob Way.
36816: sql: move gossipccl.DisableMerges to gossip, call during idx backfill r=dt a=dt This moves the DisableMerges helper from gossipccl to gossip, and adds a call to it during index backfills. Release note: none. 36832: roachtest: fix backup2TB by using IMPORT, rip out store dumps r=petermattis a=tbg Now that 19.1 is tagged, the store dumps don't work any more. Consequently the tests haven't run in over two weeks. Before that, it had episodes during which it didn't run most of the time due to flakes downloading the store dumps. Hopefully now it'll be more reliable. Rip out the unused remainders of store dumps as well. Fixes #34702. Release note: None Co-authored-by: David Taylor <tinystatemachine@gmail.com> Co-authored-by: Tobias Schottdorf <tobias.schottdorf@gmail.com>
There was a problem hiding this comment.
Reviewed 4 of 4 files at r4, 2 of 2 files at r5, 2 of 2 files at r6.
Reviewable status:
Build succeeded |
This test has been flaky because the TableDescriptor being read is being seen as OFFLINE. This issue seems to disappear after waiting a bit. This also happens during IMPORT and this issue was avoided in a similar way (see cockroachdb#36832). This should be removed and cockroachdb#40951 is tracking this issue. Release justification: Work-around for a flaky test. Release note: None
This test has been flaky because the TableDescriptor being read is being seen as OFFLINE. This issue seems to disappear after waiting a bit. This also happens during IMPORT and this issue was avoided in a similar way (see cockroachdb#36832). This should be removed and cockroachdb#40951 is tracking this issue. Release justification: Work-around for a flaky test. Release note: None
This test has been flaky because the TableDescriptor being read is being seen as OFFLINE. This issue seems to disappear after waiting a bit. This also happens during IMPORT and this issue was avoided in a similar way (see cockroachdb#36832). This should be removed and cockroachdb#40951 is tracking this issue. Release justification: Work-around for a flaky test. Release note: None
Now that 19.1 is tagged, the store dumps don't work any more. Consequently
the tests haven't run in over two weeks. Before that, it had episodes
during which it didn't run most of the time due to flakes downloading the
store dumps. Hopefully now it'll be more reliable.
Rip out the unused remainders of store dumps as well.
Fixes #34702.
Release note: None