Bug 692599

Summary:	389 DS deadlocks when mass-adding users from multiple processes
Product:	[Fedora] Fedora	Reporter:	Stephen Gallagher <sgallagh>
Component:	389-ds-base	Assignee:	Rich Megginson <rmeggins>
Status:	CLOSED DUPLICATE	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	rawhide	CC:	edewata, nalin, nhosoi, nkinder, rmeggins
Target Milestone:	---	Keywords:	screened
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:	Virtual Machine (8 CPUs) running Fedora 14 i686
Last Closed:	2011-04-12 01:04:23 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Stephen Gallagher 2011-03-31 16:40:39 UTC

Description of problem:
I needed to add fifty thousand users to my DS for testing purposes. I generated five separate LDIF files with ten thousand users in each and ran ldapmodify from separate processes on the localhost. (This was intended to take advantage of more processors, as I had eight available)

This worked for a few hundred users before 389 DS became completely unresponsive. None of the five processes reported users being added, and the error_log had nothing to say. I straced the dirsrv process and saw what appeared to be a tight-loop calling poll().

I had to kill -9 the server in order to get it back. 'service dirsrv restart' hung as well.

Version-Release number of selected component (if applicable):
389-ds-base-1.2.8-0.2.a2.fc14.i686

How reproducible:
Irregular

Steps to Reproduce:
See description
  
Actual results:
Directory server was hung.

Expected results:
All fifty thousand users should have been added

Additional info:

Comment 1 Nathan Kinder 2011-04-11 21:32:45 UTC

Stephen - Is this the same issue as bug #692690?

Comment 2 Stephen Gallagher 2011-04-12 01:04:23 UTC

Yes, I believe it is.

*** This bug has been marked as a duplicate of bug 692690 ***

Comment 3 Nalin Dahyabhai 2011-04-12 13:10:31 UTC

Are you sure?  Bug #692690 isn't a deadlock so much as a a-server-thread's-really-busy-right-now which happens to last a long time.

Comment 4 Stephen Gallagher 2011-04-12 13:15:51 UTC

Well, I suppose it could be considered a separate issue that slapi-nis taking forever in one thread is causing all other threads to block would be a separate issue.

So yeah, I guess we should reopen this.

Comment 5 Rich Megginson 2011-04-12 14:39:16 UTC

(In reply to comment #4)
> Well, I suppose it could be considered a separate issue that slapi-nis taking
> forever in one thread is causing all other threads to block would be a separate
> issue.
> 
> So yeah, I guess we should reopen this.

Does it cause all operations to block?

Comment 6 Stephen Gallagher 2011-04-12 14:40:33 UTC

(In reply to comment #5)
> (In reply to comment #4)
> > Well, I suppose it could be considered a separate issue that slapi-nis taking
> > forever in one thread is causing all other threads to block would be a separate
> > issue.
> > 
> > So yeah, I guess we should reopen this.
> 
> Does it cause all operations to block?

Yes. Once I was in the state described by bug 692690 I could not perform any search, modify, etc.