21254: Split TestRunForever into timer- and signal-triggered tests.

[arvados.git] / services / keep-balance / balance_run_test.go

diff --git a/services/keep-balance/balance_run_test.go b/services/keep-balance/balance_run_test.go
index fefd2c6c1bd440758d1d956affb8a5603140cb85..f66194e2a2cd39852d8de4cf45c8aa0a4cc89962 100644 (file)
--- a/services/keep-balance/balance_run_test.go
+++ b/services/keep-balance/balance_run_test.go
@@ -624,7 +624,7 @@ func (s *runSuite) TestChunkPrefix(c *check.C) {
        c.Check(string(lost), check.Equals, "")
 }
 
-func (s *runSuite) TestRunForever(c *check.C) {
+func (s *runSuite) TestRunForever_TriggeredByTimer(c *check.C) {
        s.config.ManagementToken = "xyzzy"
        opts := RunOptions{
                Logger: ctxlog.TestLogger(c),
@@ -640,7 +640,57 @@ func (s *runSuite) TestRunForever(c *check.C) {
 
        ctx, cancel := context.WithCancel(context.Background())
        defer cancel()
-       s.config.Collections.BalancePeriod = arvados.Duration(100 * time.Millisecond)
+       s.config.Collections.BalancePeriod = arvados.Duration(10 * time.Millisecond)
+       srv := s.newServer(&opts)
+
+       done := make(chan bool)
+       go func() {
+               srv.runForever(ctx)
+               close(done)
+       }()
+
+       // Each run should send 4 pull lists + 4 trash lists. The
+       // first run should also send 4 empty trash lists at
+       // startup. We should complete at least four runs in much less
+       // than 10s.
+       for t0 := time.Now(); time.Since(t0) < 10*time.Second; {
+               pulls := pullReqs.Count()
+               if pulls >= 16 && trashReqs.Count() == pulls+4 {
+                       break
+               }
+               time.Sleep(time.Millisecond)
+       }
+       cancel()
+       <-done
+       c.Check(pullReqs.Count() >= 16, check.Equals, true)
+       c.Check(trashReqs.Count() >= 20, check.Equals, true)
+
+       // We should have completed 4 runs before calling cancel().
+       // But the next run might also have started before we called
+       // cancel(), in which case the extra run will be included in
+       // the changeset_compute_seconds_count metric.
+       completed := pullReqs.Count() / 4
+       metrics := arvadostest.GatherMetricsAsString(srv.Metrics.reg)
+       c.Check(metrics, check.Matches, fmt.Sprintf(`(?ms).*\narvados_keepbalance_changeset_compute_seconds_count (%d|%d)\n.*`, completed, completed+1))
+}
+
+func (s *runSuite) TestRunForever_TriggeredBySignal(c *check.C) {
+       s.config.ManagementToken = "xyzzy"
+       opts := RunOptions{
+               Logger: ctxlog.TestLogger(c),
+               Dumper: ctxlog.TestLogger(c),
+       }
+       s.stub.serveCurrentUserAdmin()
+       s.stub.serveFooBarFileCollections()
+       s.stub.serveKeepServices(stubServices)
+       s.stub.serveKeepstoreMounts()
+       s.stub.serveKeepstoreIndexFoo4Bar1()
+       trashReqs := s.stub.serveKeepstoreTrash()
+       pullReqs := s.stub.serveKeepstorePull()
+
+       ctx, cancel := context.WithCancel(context.Background())
+       defer cancel()
+       s.config.Collections.BalancePeriod = arvados.Duration(time.Minute)
        srv := s.newServer(&opts)
 
        done := make(chan bool)
@@ -654,36 +704,29 @@ func (s *runSuite) TestRunForever(c *check.C) {
 
        // Each run should send 4 pull lists + 4 trash lists. The
        // first run should also send 4 empty trash lists at
-       // startup. We should complete all four runs in much less than
-       // a second.
+       // startup. We should be able to complete four runs in much
+       // less than 10s.
        completedRuns := 0
        for t0 := time.Now(); time.Since(t0) < 10*time.Second; {
                pulls := pullReqs.Count()
                if pulls >= 16 && trashReqs.Count() == pulls+4 {
                        break
                }
-               if pulls > 4 {
-                       // Once the 2nd run has started automatically
-                       // (indicating that our BalancePeriod is
-                       // working) we switch to a long wait time to
-                       // effectively stop the timed runs, and
-                       // instead start sending a single SIGUSR1 at
-                       // the end of each (2nd or 3rd) run, to ensure
-                       // we get exactly 4 runs in total.
-                       srv.Cluster.Collections.BalancePeriod = arvados.Duration(time.Minute)
-                       if pulls%4 == 0 && pulls <= 12 && pulls/4 > completedRuns {
-                               completedRuns = pulls / 4
-                               c.Logf("completed run %d, sending SIGUSR1 to trigger next run", completedRuns)
-                               procself.Signal(syscall.SIGUSR1)
-                       }
+               // Once the 1st run has started automatically, we
+               // start sending a single SIGUSR1 at the end of each
+               // run, to ensure we get exactly 4 runs in total.
+               if pulls > 0 && pulls%4 == 0 && pulls <= 12 && pulls/4 > completedRuns {
+                       completedRuns = pulls / 4
+                       c.Logf("completed run %d, sending SIGUSR1 to trigger next run", completedRuns)
+                       procself.Signal(syscall.SIGUSR1)
                }
                time.Sleep(time.Millisecond)
        }
        cancel()
        <-done
-       c.Check(pullReqs.Count() >= 16, check.Equals, true)
-       c.Check(trashReqs.Count(), check.Equals, pullReqs.Count()+4)
+       c.Check(pullReqs.Count(), check.Equals, 16)
+       c.Check(trashReqs.Count(), check.Equals, 20)
 
        metrics := arvadostest.GatherMetricsAsString(srv.Metrics.reg)
-       c.Check(metrics, check.Matches, `(?ms).*\narvados_keepbalance_changeset_compute_seconds_count `+fmt.Sprintf("%d", pullReqs.Count()/4)+`\n.*`)
+       c.Check(metrics, check.Matches, `(?ms).*\narvados_keepbalance_changeset_compute_seconds_count 4\n.*`)
 }