On 05/10/2011 08:34 AM, Reinhard Nappert wrote:
Rick,
 
it seems that the issue does not exist in 1.2.8.2. I just compiled and installed it.
 
If 1.2.8.2 is stable, I won't bother open a bug for it......
Ok, thanks for confirming.  1.2.8.3 is in Testing now - I'm going to push that to Stable in the next couple of days.
 
-Reinhard


From: Rich Megginson [mailto:rmeggins@redhat.com]
Sent: Monday, May 09, 2011 9:37 PM
To: General discussion list for the 389 Directory server project.
Cc: Reinhard Nappert
Subject: Re: [389-users] MMR issue, when deleting the replica setup.

On 05/09/2011 02:06 PM, Reinhard Nappert wrote:
Hi,
 
I noticed an issue with 389 DS 1.2.7.5, which I have not seen before. Here is what I do:
1. I create a two multi-master setup.
2. I don't perform any changes on the directory.
3. I delete the replica setup on both systems -- everything is fine.
4. I create a two multi-master setup.
5. Perform changes on both systems
6. Modifications get replicated.
7 I delete the replica setup. No I get the following error logs:
 
[09/May/2011:15:43:18 -0400] - import userRoot: Import complete.  Processed 446 entries in 4 seconds. (111.50 entries/sec)
[09/May/2011:15:43:18 -0400] NSMMReplicationPlugin - multimaster_be_state_change: replica o=base is coming online; enabling replication
...
[09/May/2011:15:45:21 -0400] NSMMReplicationPlugin - agmt_delete: begin
[09/May/2011:15:45:22 -0400] NSMMReplicationPlugin - replica_config_delete: Warning: The changelog for replica o=BASE is no longer valid since the replica config is being deleted.  Removing the changelog.
[09/May/2011:15:45:22 -0400] NSMMReplicationPlugin - changelog program - _cl5Add Thread: invalid changelog state - 2 <== This is good!
[09/May/2011:15:45:27 -0400] - libdb: <path to>/changelogdb/7773fd02-7a7411e0-ac71f4b1-0fb2d026_4dc840d3000000020000.db4: unable to flush: No such file or directory
[09/May/2011:15:45:27 -0400] - libdb: txn_checkpoint: failed to flush the buffer cache No such file or directory
[09/May/2011:15:45:27 -0400] - Serious Error---Failed to checkpoint database, err=2 (No such file or directory)
Of course, the changelog directory was gone. It looks to me that the server keeps this still somehow in memory.
 
I enabled the audit-logging: This is what I see there:
 
time: 20110509154521
dn: cn=changelog5,cn=config
changetype: delete
modifiersname: <credentials>
time: 20110509154522
dn: cn=agreement1,cn=replica,cn=o\3dbase,cn=mapping tree,cn=config
changetype: delete
modifiersname: <credentials>
 
time: 20110509154522
dn: cn=replica,cn=o\3dbase,cn=mapping tree,cn=config
changetype: delete
modifiersname: <credentials>
time: 20110509154522
dn: cn=o\3dbase,cn=mapping tree,cn=config
changetype: modify
replace: nsslapd-state
nsslapd-state: backend
-
replace: nsslapd-referral
-
replace: modifiersname
modifiersname: <credentials>-
replace: modifytimestamp
-
replace: nsslapd-referral
-
replace: modifiersname
modifiersname: <credentials>
-
replace: modifytimestamp
modifytimestamp: 20110509194522Z
-
 
time: 20110509154605
dn: cn=uniqueid generator,cn=config
changetype: modify
replace: nsState
nsState:: AM+94nR64AH0sQ+y0CZxbAEAAAAAAAAA
-
replace: modifiersname
modifiersname: cn=server,cn=plugins,cn=config
-
replace: modifytimestamp
modifytimestamp: 20110509194605Z
-
Has somebody has seen this before.
No, please file a bug.
 
Thanks,
-Reinhard
-- 389 users mailing list 389-users@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/389-users