Project

General

Profile

Actions
Copy link

Feature #10037

closed

cache-tier: Optimise RBD image removal

Added by Dmitry Smirnov over 9 years ago. Updated almost 7 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Target version:
-
% Done:

0%

Source:
Community (user)
Tags:
Backport:
Reviewed:
Affected Versions:
Pull request ID:

Description

While removing an RBD image from EC pool I've noticed that it bubbles-up to caching pool hence removal is very slow. It seems unnecessary to read all RBD image data from EC pool and propagate it through caching pool on deletion. Not only it is slow by itself but also it unnecessarily moves data which interferes with other clients. Please optimise.

Actions
Copy link
 #1

Updated by Jason Dillaman almost 7 years ago

  • Subject changed from Optimise RBD image removal to cache-tier: Optimise RBD image removal
Actions
Copy link
 #2

Updated by Greg Farnum almost 7 years ago

  • Project changed from Ceph to rbd
  • Category deleted (librbd)

I think the proxy changes have fixed this for deletes on the RADOS side; does rbd do anything which would force promote or is it done?

Actions
Copy link
 #3

Updated by Jason Dillaman almost 7 years ago

  • Status changed from New to Resolved

RBD only issues remove ops against all possible objects -- and with object map enabled it only issues them against objects it knows exist. Assuming the remove op is write-proxied to the base tier, this should be resolved.

Actions
Copy link

Also available in: Atom PDF