Project

General

Profile

Feature #58550

Feature #58488: mds: avoid encoding srnode for each ancestor in an EMetaBlob log event

mds: add perf counter to track (relatively) larger log events

Added by Venky Shankar 8 months ago. Updated about 2 months ago.

Status:
Resolved
Priority:
Normal
Assignee:
Patrick Donnelly
Category:
Administration/Usability
Target version:
Ceph - v19.0.0
% Done:

0%

Source:
Development
Tags:
Backport:
Reviewed:
Affected Versions:
Component(FS):
MDS
Labels (FS):
Pull request ID:
48732

Description

Logging of large (in size) subtreemap over and over again in the log segment is a well know scalability limiting factor in the MDS. Its a bit hard to infer this though when a cluster starts experiencing slowness. Moreover, there might be other places where a large log event size could be making things worse. Adding a perf counter that increments whenever the log event size exceeds a certain threshold might help is inferring the extent of logging large log events over and over again.

History

#1 Updated by Venky Shankar 8 months ago

  • Category set to Administration/Usability
  • Assignee set to Patrick Donnelly

#2 Updated by Patrick Donnelly 8 months ago

  • Status changed from New to In Progress

#3 Updated by Patrick Donnelly 8 months ago

  • Status changed from In Progress to Fix Under Review
  • Pull request ID set to 48732

#4 Updated by Patrick Donnelly about 2 months ago

  • Status changed from Fix Under Review to Resolved
  • Target version changed from v18.0.0 to v19.0.0
  • Source set to Development
  • Backport deleted (pacific,quincy)

Not planning to backport this the minor log segment PR.

Also available in: Atom PDF