Project

General

Profile

Actions

Bug #4567

closed

[Crunch] job dies but is left in state "locked"

Added by Tim Pierce about 10 years ago. Updated about 10 years ago.

Status:
Closed
Priority:
Normal
Assigned To:
-
Category:
Crunch
Target version:
-
Start date:
11/17/2014
Due date:
% Done:

0%

Estimated time:
Story points:
-

Description

Job https://workbench.9tee4.arvadosapi.com/jobs/9tee4-8i9sb-q2qexkqhszi80pv died earlier today -- it was not running on either of the live compute nodes when I checked, but was still marked in state "running."

When I pressed "Cancel", I got a fiddlesticks page with error token 1416267205+1fc8d6cd, which mapped to this error in the log:

User 9tee4-tpzed-fdwy19r69r16594 tried to change protected job attributes on locked Job 9tee4-8i9sb-q2qexkqhszi80pv

Related issues 2 (1 open1 closed)

Is duplicate of Arvados - Bug #4303: [Crunch] Output of job gets stuck at collate stepResolvedBrett Smith10/31/2014

Actions
Is duplicate of Arvados - Bug #4273: [API] as admin user, permission denied canceling a job (started by another user)New

Actions
Actions #1

Updated by Tim Pierce about 10 years ago

  • Category set to Crunch
  • Target version set to Bug Triage
Actions #2

Updated by Brett Smith about 10 years ago

The job hung because it was trying and failing to record its output. Eventually Crunch reported: "Failed to register output manifest: API call /collections failed: 502 Bad Gateway" #4303 makes it clearer in the logs when this is happening. That wasn't deployed at the time the job ran.

The second half of this bug is #4273.

Closing as a duplicate of both of these.

Actions #3

Updated by Brett Smith about 10 years ago

  • Status changed from New to Closed
Actions #4

Updated by Brett Smith about 10 years ago

  • Target version deleted (Bug Triage)
Actions

Also available in: Atom PDF