Project

General

Profile

Bug #19236

multisite: some 'radosgw-admin data sync' commands hang

Added by Casey Bodley 5 months ago. Updated 4 months ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Target version:
-
Start date:
03/08/2017
Due date:
% Done:

0%

Source:
Tags:
Backport:
jewel, kraken
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Release:
Needs Doc:
No

Description

RGWRemoteDataLog::init_sync_status() and read_sync_status() use their own RGWCoroutinesManager, because RGWCoroutinesManager::run() is not reentrant. The rest api needs to call these functions while the sync thread is in RGWRemoteDataLog::run_sync().

However, those coroutines were still using RGWRemoteDataLog::http_manager, which was sending completions to RGWRemoteDataLog's RGWCoroutinesManager. This prevented the coroutines from waking up when the replies came back, and results in hangs.

https://github.com/ceph/ceph/pull/13660


Related issues

Copied to rgw - Backport #19353: jewel: multisite: some 'radosgw-admin data sync' commands hang Resolved
Copied to rgw - Backport #19354: kraken: multisite: some 'radosgw-admin data sync' commands hang Resolved

History

#1 Updated by Casey Bodley 5 months ago

  • Backport set to jewel kraken

#2 Updated by Orit Wasserman 5 months ago

  • Status changed from Need Review to Pending Backport

#3 Updated by Nathan Cutler 5 months ago

  • Copied to Backport #19353: jewel: multisite: some 'radosgw-admin data sync' commands hang added

#4 Updated by Nathan Cutler 5 months ago

  • Copied to Backport #19354: kraken: multisite: some 'radosgw-admin data sync' commands hang added

#5 Updated by Nathan Cutler 4 months ago

  • Status changed from Pending Backport to Resolved
  • Backport changed from jewel kraken to jewel, kraken

Also available in: Atom PDF