Bug #1357

Making many pools and writing to them eventually hangs

Added by Greg Farnum almost 2 years ago. Updated almost 2 years ago.

Status:ResolvedStart date:08/04/2011
Priority:NormalDue date:
Assignee:Greg Farnum% Done:

0%

Category:OSDSpent time:10.50 hours
Target version:v0.33
Source: Severity:
Backport: Reviewed:
Tags:

Description

Discovered while testing #1297. Looping pool creates eventually hang when the OSD rejects writes for being sent earlier than same_primary_since on the PG.
Turns out that the OSD was initializing same_primary_since based on the osdmap epoch when the PG was created locally, not when it was entered into the map.

A Fix is in testing now.

History

#1 Updated by Greg Farnum almost 2 years ago

  • Status changed from New to Testing

#2 Updated by Greg Farnum almost 2 years ago

  • Status changed from Testing to In Progress

Well, I've found several different races and bugs. My current progress is in the wip-pg-creation branch, but there's still at least one issue creating...
Hopefully I can track it down tomorrow or on my way home.

#3 Updated by Greg Farnum almost 2 years ago

  • Status changed from In Progress to Resolved

Ah, I didn't have quite the right fix in for the PGMonitor's PG creation (I was only setting it properly for the localized PGs -- whoops!).

This issue is now resolved. However, there is a similar one that crops up when creating new PGs in a pre-existing pool. I've made bug #1358 to track that.

Also available in: Atom PDF