Project

General

Profile

Actions

Bug #8043

closed

until we fix it more better, we should disallow split on cache pools

Added by Samuel Just about 10 years ago. Updated about 10 years ago.

Status:
Resolved
Priority:
Urgent
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Source:
Development
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Actions #1

Updated by Sage Weil about 10 years ago

  • Source changed from other to Development

Is the concern the need for scrub, or are there other issues?

We could make the (shallow) scrub triggered automatically when the agent sees the invalid_stats flag. Or, we could make the mon refuse to split unless provided a --force flag...

Actions #2

Updated by Greg Farnum about 10 years ago

Requiring a scrub first means that any cache pool which is actually being used and gets split will fill up and block IO (because it will be running nearly full to begin with) before the scrub completes.

Actions #3

Updated by Sage Weil about 10 years ago

How about a flag in the pg info that prevents the agent from blocking on full until it sees the thresholds drop below the target?

Actions #4

Updated by Greg Farnum about 10 years ago

Do you really expect cache pool targets to not be set based on the underlying storage capacity? (Hint: no.)

;)

Actions #5

Updated by Sage Weil about 10 years ago

  • Status changed from New to Fix Under Review
Actions #6

Updated by Sage Weil about 10 years ago

  • Status changed from Fix Under Review to Resolved
Actions

Also available in: Atom PDF