Skip to content

Commit

Permalink
roachtest: skip bank/zerosum-restart
Browse files Browse the repository at this point in the history
It flakes on various relocate errors that are "expected" when nodes
are freshly restarted due to Gossip asynchronicity. However we've
decided to work on that root cause instead of adding more random
exceptions to the test. As a result, we're skipping it for now.

Release note: None
  • Loading branch information
tbg committed Jan 16, 2019
1 parent 0c7e491 commit 7160776
Show file tree
Hide file tree
Showing 2 changed files with 13 additions and 2 deletions.
6 changes: 5 additions & 1 deletion pkg/cmd/roachtest/acceptance.go
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,10 @@ func registerAcceptance(r *registry) {
// local mode the acceptance tests should be configured to run within a
// minute or so as these tests are run on every merge to master.

// NB: zerosum-restart is skipped due to generating various errors during
// its rebalances. See the comment on:
_ = isExpectedRelocateError

testCases := []struct {
name string
fn func(ctx context.Context, t *test, c *cluster)
Expand All @@ -33,7 +37,7 @@ func registerAcceptance(r *registry) {
{"bank/cluster-recovery", runBankClusterRecovery},
{"bank/node-restart", runBankNodeRestart},
{"bank/zerosum-splits", runBankNodeZeroSum},
{"bank/zerosum-restart", runBankZeroSumRestart},
// {"bank/zerosum-restart", runBankZeroSumRestart},
{"build-info", runBuildInfo},
{"cli/node-status", runCLINodeStatus},
{"decommission", runDecommissionAcceptance},
Expand Down
9 changes: 8 additions & 1 deletion pkg/cmd/roachtest/bank.go
Original file line number Diff line number Diff line change
Expand Up @@ -345,7 +345,14 @@ func (s *bankState) startSplitMonkey(ctx context.Context, d time.Duration, c *cl
}

func isExpectedRelocateError(err error) bool {
return testutils.IsError(err, "(descriptor changed|unable to remove replica .* which is not present|unable to add replica .* which is already present|received invalid ChangeReplicasTrigger .* to remove self|breaker open)")
// See:
// https://github.com/cockroachdb/cockroach/issues/33732
// https://github.com/cockroachdb/cockroach/issues/33708
// https://github.cm/cockroachdb/cockroach/issues/34012
// for more failure modes not caught here. We decided to avoid adding
// to this catchall and to fix the root causes instead.
// We've also seen "breaker open" errors here.
return testutils.IsError(err, "(descriptor changed|unable to remove replica .* which is not present|unable to add replica .* which is already present|received invalid ChangeReplicasTrigger .* to remove self)")
}

func accountDistribution(r *rand.Rand) *rand.Zipf {
Expand Down

0 comments on commit 7160776

Please sign in to comment.