X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/3aaefcb3c76ff470b475d950398d01255e87712a..8288bba27e9beff7273aeb65c5200248e52bab02:/lib/dispatchcloud/worker/worker.go

diff --git a/lib/dispatchcloud/worker/worker.go b/lib/dispatchcloud/worker/worker.go
index 9e89d7daaf..b01a820cd6 100644
--- a/lib/dispatchcloud/worker/worker.go
+++ b/lib/dispatchcloud/worker/worker.go
@@ -313,6 +313,10 @@ func (wkr *worker) probeAndUpdate() {
 		// not yet running when ctrUUIDs was generated. Leave
 		// wkr.running alone and wait for the next probe to
 		// catch up on any changes.
+		logger.WithFields(logrus.Fields{
+			"updated":     updated,
+			"wkr.updated": wkr.updated,
+		}).Debug("skipping worker state update due to probe/sync race")
 		return
 	}
 
@@ -387,6 +391,11 @@ func (wkr *worker) probeRunning() (running []string, reportsBroken, ok bool) {
 		wkr.wp.mRunProbeDuration.WithLabelValues("fail").Observe(time.Now().Sub(before).Seconds())
 		return
 	}
+	wkr.logger.WithFields(logrus.Fields{
+		"Command": cmd,
+		"stdout":  string(stdout),
+		"stderr":  string(stderr),
+	}).Debug("probe succeeded")
 	wkr.wp.mRunProbeDuration.WithLabelValues("success").Observe(time.Now().Sub(before).Seconds())
 	ok = true
 
@@ -409,6 +418,12 @@ func (wkr *worker) probeRunning() (running []string, reportsBroken, ok bool) {
 			// empty string following final newline
 		} else if s == "broken" {
 			reportsBroken = true
+		} else if !strings.HasPrefix(s, wkr.wp.cluster.ClusterID) {
+			// Ignore crunch-run processes that belong to
+			// a different cluster (e.g., a single host
+			// running multiple clusters with the loopback
+			// driver)
+			continue
 		} else if toks := strings.Split(s, " "); len(toks) == 1 {
 			running = append(running, s)
 		} else if toks[1] == "stale" {