Merge branch '9766-register-workflow' closes #9766

[arvados.git] / services / crunch-dispatch-slurm / crunch-dispatch-slurm.go

diff --git a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
index ef08f4583a6232e245a8375602c8b438bcb5a368..b33dc64e7bd7bf7b854fde1a55d0a388b32407db 100644 (file)
--- a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
+++ b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
@@ -151,10 +151,7 @@ func submit(dispatcher *dispatch.Dispatcher,
                        // OK, no cleanup needed
                        return
                }
-               err := dispatcher.Arv.Update("containers", container.UUID,
-                       arvadosclient.Dict{
-                               "container": arvadosclient.Dict{"state": "Queued"}},
-                       nil)
+               err := dispatcher.Unlock(container.UUID)
                if err != nil {
                        log.Printf("Error unlocking container %s: %v", container.UUID, err)
                }
@@ -247,7 +244,7 @@ func monitorSubmitOrCancel(dispatcher *dispatch.Dispatcher, container arvados.Co
                                log.Printf("Error submitting container %s to slurm: %v",
                                        container.UUID, err)
                                // maybe sbatch is broken, put it back to queued
-                               dispatcher.UpdateState(container.UUID, dispatch.Queued)
+                               dispatcher.Unlock(container.UUID)
                        }
                        submitted = true
                } else {
@@ -263,20 +260,20 @@ func monitorSubmitOrCancel(dispatcher *dispatch.Dispatcher, container arvados.Co
                                log.Printf("Error getting final container state: %v", err)
                        }
 
-                       var st arvados.ContainerState
                        switch con.State {
                        case dispatch.Locked:
-                               st = dispatch.Queued
+                               log.Printf("Container %s in state %v but missing from slurm queue, changing to %v.",
+                                       container.UUID, con.State, dispatch.Queued)
+                               dispatcher.Unlock(container.UUID)
                        case dispatch.Running:
-                               st = dispatch.Cancelled
+                               st := dispatch.Cancelled
+                               log.Printf("Container %s in state %v but missing from slurm queue, changing to %v.",
+                                       container.UUID, con.State, st)
+                               dispatcher.UpdateState(container.UUID, st)
                        default:
                                // Container state is Queued, Complete or Cancelled so stop monitoring it.
                                return
                        }
-
-                       log.Printf("Container %s in state %v but missing from slurm queue, changing to %v.",
-                               container.UUID, con.State, st)
-                       dispatcher.UpdateState(container.UUID, st)
                }
        }
 }