From dd576c00f842e386c7d7b184f77bb085fdb853c9 Mon Sep 17 00:00:00 2001 From: Tobias Grieger Date: Thu, 27 Aug 2020 11:26:39 +0200 Subject: [PATCH 1/4] roachtest: own disk-stalled roachtest to storage team Release justification: testing-only Release note: None --- pkg/cmd/roachtest/disk_stall.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/pkg/cmd/roachtest/disk_stall.go b/pkg/cmd/roachtest/disk_stall.go index 2e151b42a196..2d5a65b95b57 100644 --- a/pkg/cmd/roachtest/disk_stall.go +++ b/pkg/cmd/roachtest/disk_stall.go @@ -33,7 +33,7 @@ func registerDiskStalledDetection(r *testRegistry) { "disk-stalled/log=%t,data=%t", affectsLogDir, affectsDataDir, ), - Owner: OwnerKV, + Owner: OwnerStorage, MinVersion: "v19.2.0", Cluster: makeClusterSpec(1), Run: func(ctx context.Context, t *test, c *cluster) { From 286e3c009b977c831bd08f51efa2af0df637d157 Mon Sep 17 00:00:00 2001 From: Tobias Grieger Date: Thu, 27 Aug 2020 11:39:20 +0200 Subject: [PATCH 2/4] roachtest: require 20.1+ for some tpccbench versions Don't run `tpccbench/nodes=9/cpu=4/`chaos/partition,/multi-region}` no 19.2 and older. These tests are hard to stabilize and we are unlikely to have the bandwidth to do so on old release branches. Release justification: testing change Release note: None --- pkg/cmd/roachtest/tpcc.go | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/pkg/cmd/roachtest/tpcc.go b/pkg/cmd/roachtest/tpcc.go index b82ae6431227..bc42e55aa3d8 100644 --- a/pkg/cmd/roachtest/tpcc.go +++ b/pkg/cmd/roachtest/tpcc.go @@ -414,7 +414,7 @@ func registerTPCC(r *testRegistry) { LoadWarehouses: 5000, EstimatedMax: 3000, - MinVersion: "v19.1.0", + MinVersion: "v20.1.0", }) registerTPCCBenchSpec(r, tpccBenchSpec{ Nodes: 9, @@ -425,7 +425,7 @@ func registerTPCC(r *testRegistry) { LoadWarehouses: 2000, EstimatedMax: 900, - MinVersion: "v19.1.0", + MinVersion: "v20.1.0", }) } From 38af67605829d3eb2a5725cb10b1e59595d2eff2 Mon Sep 17 00:00:00 2001 From: Tobias Grieger Date: Thu, 27 Aug 2020 11:54:24 +0200 Subject: [PATCH 3/4] roachtest: run jepsen on 20.1+ only It generates a fair amount of flakes and we're unlikely to land any changes on these old branches that can expose new behavior. Release justification: testing Release note: None --- pkg/cmd/roachtest/jepsen.go | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/pkg/cmd/roachtest/jepsen.go b/pkg/cmd/roachtest/jepsen.go index 871fc85f5311..a8fc3c003457 100644 --- a/pkg/cmd/roachtest/jepsen.go +++ b/pkg/cmd/roachtest/jepsen.go @@ -312,8 +312,10 @@ func registerJepsen(r *testRegistry) { for _, nemesis := range jepsenNemeses { nemesis := nemesis // copy for closure spec := testSpec{ - Name: fmt.Sprintf("jepsen/%s/%s", testName, nemesis.name), - Owner: OwnerKV, + Name: fmt.Sprintf("jepsen/%s/%s", testName, nemesis.name), + // We don't run jepsen on older releases due to the high rate of flakes. + MinVersion: "v20.1.0", + Owner: OwnerKV, // The Jepsen tests do funky things to machines, like muck with the // system clock; therefore, their clusters cannot be reused other tests // except the Jepsen ones themselves which reset all this state when From 5343d568ea7e37246574b1938a642e4abbe667d5 Mon Sep 17 00:00:00 2001 From: Tobias Grieger Date: Thu, 27 Aug 2020 13:08:13 +0200 Subject: [PATCH 4/4] roachtest: de-flake gossip/peerings Pulling the gossip info via HTTP can fail if the freshly restarted nodes didn't gossip the actual cluster settings yet. Fixing the initialization is tracked via #50271. Closes #48005. Release justification: testing Release note: None --- pkg/cmd/roachtest/gossip.go | 3 +++ 1 file changed, 3 insertions(+) diff --git a/pkg/cmd/roachtest/gossip.go b/pkg/cmd/roachtest/gossip.go index 1df5077c3965..988e36754f31 100644 --- a/pkg/cmd/roachtest/gossip.go +++ b/pkg/cmd/roachtest/gossip.go @@ -271,6 +271,9 @@ func runGossipPeerings(ctx context.Context, t *test, c *cluster) { t.l.Printf("%d: restarting node %d\n", i, node[0]) c.Stop(ctx, node) c.Start(ctx, t, node) + // Sleep a bit to avoid hitting: + // https://github.com/cockroachdb/cockroach/issues/48005 + time.Sleep(3 * time.Second) } }