Project

General

Profile

Actions

Bug #6306

closed

[Workbench] Indicate that job progress/state is out-of-date when showing pipeline instance

Added by Bryan Cosca almost 9 years ago. Updated over 4 years ago.

Status:
Closed
Priority:
Normal
Assigned To:
-
Category:
Workbench
Target version:
-
Story points:
0.5

Description

summary

When a pipeline fails, it stops updating its own cache of job status. However, some jobs may continue to run. In such cases, Workbench displays either up-to-date job status or out-of-date job status, depending on whether the current user has permission to look up the jobs themselves. This is confusing.

implementation

When displaying a pipeline instance that is not Running, if the cached job state is not already Failed or Complete, and Workbench fails to look up the current state, it should
  • show a "current job state not available" message
  • fade out or completely hide the cached state

initial bug report

on my admin account where I ran the pipeline in the home directory, I have https://cloud.curoverse.com/pipeline_instances/qr1hi-d1hrv-pkxdm18hmf8jsht (bcosc@curoverse)

I copied it over to the Making BWA faster project and when I went on my non-admin account: () I have
https://cloud.curoverse.com/pipeline_instances/qr1hi-d1hrv-pkxdm18hmf8jsht

when I click on the instance from , the time that samtools-merge is running for changes but not to the time that I waited to click on it. Also, there are no busy nodes, so it doesn't seem like it is actually running.


Files

instance_nosync.png (67.2 KB) instance_nosync.png Bryan Cosca, 06/11/2015 02:25 PM
instance_nosync2.png (61 KB) instance_nosync2.png Bryan Cosca, 06/11/2015 02:25 PM
Actions #1

Updated by Bryan Cosca almost 9 years ago

  • Description updated (diff)
Actions #2

Updated by Tom Clegg almost 9 years ago

Each pipeline instance has a cache of the last known state of each component job, in addition to the jobs' UUIDs.

Once a pipeline instance fails, it stops updating its cache. When Workbench displays the instance, it looks up the job UUIDs to get the real current state, so you don't notice that the cache has stopped updating.

However, if you don't have permission to see those jobs, the "real current job state" lookup fails, and Workbench just displays the cached state.

If we set aside (at least for now) the question of managing permissions so this doesn't happen by surprise, I think the solution is for Workbench to display an error condition ("{!} job detail not available"??) when the cached job state isn't final (i.e., state isn't Failed or Complete) and it fails to look up the current job state.

Actions #3

Updated by Tom Clegg almost 9 years ago

  • Subject changed from Copying a pipeline instance did not sync with another user to [Workbench] Indicate that job progress/state is out-of-date when showing pipeline instance
  • Description updated (diff)
Actions #4

Updated by Tom Clegg almost 9 years ago

  • Category set to Workbench
  • Story points set to 0.5
Actions #5

Updated by Peter Amstutz over 4 years ago

  • Status changed from New to Closed
Actions

Also available in: Atom PDF