Ticket #275 (closed defect: fixed)

Opened 6 years ago

Last modified 6 years ago

Mismatch between AM and SM/Broker in accounting for VMs that have failed to come up

Reported by: ibaldin Owned by: ibaldin
Priority: major Milestone:
Component: ORCA: Policies and their components Version: baseline
Keywords: Cc: yxin@…

Description

We keep getting into situations where the broker thinks cores are available, but the AM doesn't (returns an error that indicates cores aren't available). Is this a bug in the accounting? Perhaps in modify? Need to check and test.

Please annotate this ticket with findings.

Change History

Changed 6 years ago by ibaldin

When ImagePRoxy fails to retrieve the image from a URL this results in a strange condition, where the AM thinks the reservation is active, while the SM thinks it failed:

pequod:show>show reservations for a4d85af0-ebcd-4d89-b840-c089d27eacff actor topology-embed-service
29eec281-849c-48b9-adb8-6b3a15d14189 topology-embed-service

Slice: a4d85af0-ebcd-4d89-b840-c089d27eacff
1 rcivmsite.vm [ failed, nascent]
Notices: Reservation 29eec281-849c-48b9-adb8-6b3a15d14189 (Slice test-fail-rc2) is in state [Failed,None], err=resources failed to join: (no details)
Start: Wed Jul 17 10:44:36 EDT 2013 End:Thu Jul 18 10:44:37 EDT 2013

Total: 1 reservations
pequod:show>show reservations for a4d85af0-ebcd-4d89-b840-c089d27eacff actor rci-vm-am
29eec281-849c-48b9-adb8-6b3a15d14189 rci-vm-am

Slice: a4d85af0-ebcd-4d89-b840-c089d27eacff
0 rcivmsite.vm [ active, nascent]
Notices: Reservation 29eec281-849c-48b9-adb8-6b3a15d14189 (Slice test-fail-rc2) is in state [Active,None]

Exception during join for unit: 10DED966 ImageProxy? unable to retrieve image: org.apache.axis2.AxisFault?: Error encountered while attempting to establish HTTP connection to URL: http://geni-images.renci.org/images/standard/debian/de-neuca-v1.0.7.xml ; reason was: http://geni-images.renci.org/images/standard/debian/de-neuca-v1.0.7.xml
null

Start: Wed Jul 17 10:44:36 EDT 2013 End:Thu Jul 18 10:44:37 EDT 2013

Total: 1 reservations
pequod:show>

Asked Aydan for clarification.

Changed 6 years ago by ibaldin

  • summary changed from Check core accounting (especially with modify) to Mismatch between AM and SM/Broker in accounting for VMs that have failed to come up

Changed 6 years ago by ibaldin

  • owner changed from anirban to ibaldin
  • status changed from new to assigned

Changed 6 years ago by ibaldin

  • status changed from assigned to closed
  • resolution set to fixed

Fixed in #5700

Changed 6 years ago by ibaldin

Fixed in r5700. Caveat due to udd and errmsg in AuthorityReservation? and Reservation keeping duplicate information about errors in some cases, the AM ends up printing the error twice. On SM it works fine as it comes only from UDD.

Note: See TracTickets for help on using tickets.