From: Tom Clegg <tom@tomclegg.ca>
Date: Mon, 31 Aug 2020 14:30:04 +0000 (-0400)
Subject: Merge branch '16723-kill-vs-requeue'
X-Git-Tag: 2.1.0~97
X-Git-Url: https://git.arvados.org/arvados.git/commitdiff_plain/f3e3a6cc4b72120f856e14f3039b1c0c1f0326bd?hp=5ce6d49b5a632b8e846fb0d794444ecd39f74fa5

Merge branch '16723-kill-vs-requeue'

closes #16723

Arvados-DCO-1.1-Signed-off-by: Tom Clegg <tom@tomclegg.ca>
---

diff --git a/lib/dispatchcloud/scheduler/run_queue.go b/lib/dispatchcloud/scheduler/run_queue.go
index dddb974b32..d77dcee947 100644
--- a/lib/dispatchcloud/scheduler/run_queue.go
+++ b/lib/dispatchcloud/scheduler/run_queue.go
@@ -51,6 +51,10 @@ tryrun:
 				overquota = sorted[i:]
 				break tryrun
 			}
+			if sch.pool.KillContainer(ctr.UUID, "about to lock") {
+				logger.Info("not locking: crunch-run process from previous attempt has not exited")
+				continue
+			}
 			go sch.lockContainer(logger, ctr.UUID)
 			unalloc[it]--
 		case arvados.ContainerStateLocked:
@@ -88,7 +92,7 @@ tryrun:
 				// a higher-priority container on the
 				// same instance type. Don't let this
 				// one sneak in ahead of it.
-			} else if sch.pool.KillContainer(ctr.UUID, "about to lock") {
+			} else if sch.pool.KillContainer(ctr.UUID, "about to start") {
 				logger.Info("not restarting yet: crunch-run process from previous attempt has not exited")
 			} else if sch.pool.StartContainer(it, ctr) {
 				// Success.