Bug #5917
closed
Node Manager did not spin up enough nodes when I set min_nodes = 20
Added by Bryan Cosca almost 10 years ago.
Updated almost 10 years ago.
Description

Job qr1hi-8i9sb-6hfkeo7gvpexcjw has been queued for 6.5 hours. When I logged in last night, there were 18 nodes idle and the job still did not run.
Files
- Status changed from New to Closed
Closing as a sort of duplicate. This situation is caused by two bugs we already have filed, #4129 and #5824. #5824 causes 20-node jobs to use more than half of the nodes that Arvados clusters are currently configured to handle. With all that room taken up, #4129 gets us into a situation where we can't create more nodes. Both need to be fixed. #5824 is a high priority right now because it's costing us money.
Also available in: Atom
PDF