Project

General

Profile

Bug #41198

Resolved a problem where too many requests on an object caused OSD processing suicide over time

Added by 侯 斌 13 days ago. Updated 13 days ago.

Status:
Need Review
Priority:
Normal
Assignee:
-
Category:
-
Target version:
Start date:
08/12/2019
Due date:
08/13/2019
% Done:

60%

Estimated time:
12.00 h
Source:
Community (dev)
Tags:
ObjecterCacher
Backport:
Mimic,Luminous,Nautilus
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
fs
Component(RADOS):
Pull request ID:

Description

When there is a large number of skip writes on an object,and each time only a few bytes are written, the OC will send an operation containing tens of thousands of write requests, causing OSD processing to time out.

So,i think wo can add an configuration item ,When the number of requests for an object on the OC reaches 1000,We force this object to be flushed from the OC.

History

#1 Updated by Neha Ojha 13 days ago

  • Project changed from Ceph to RADOS
  • Category deleted (ObjectCacher)
  • Status changed from New to Need Review
  • Pull request ID set to 29601

Also available in: Atom PDF