Incremental Replication over SSL ( and startTLS) with simple bind crashes the latest version

[Andrey Ivanov]

I wanted to configure the replication over SSL (both with SSL
mechanism which was available in previous versions) and by TLS using
simple bind (both in multimaster or single master-dedicated consumer models).

I've tried to configure it with command line and with the console. The
configuration and the initial initialisation are ok :

[25/Aug/2010:18:30:44 +0200] NSMMReplicationPlugin - replica_config_delete: Warning: The
changelog for replica dc=id,dc=polytechnique,dc=edu is no longer valid since the replica
config is being deleted.  Removing the changelog.
[25/Aug/2010:18:34:33 +0200] NSMMReplicationPlugin - multimaster_be_state_change: replica
dc=id,dc=polytechnique,dc=edu is going offline; disabling replication
[25/Aug/2010:18:34:33 +0200] - WARNING: Import is running with
nsslapd-db-private-import-mem on; No other process is allowed to access the database
[25/Aug/2010:18:34:39 +0200] - import userRoot: Workers finished; cleaning up...
[25/Aug/2010:18:34:40 +0200] - import userRoot: Workers cleaned up.
[25/Aug/2010:18:34:40 +0200] - import userRoot: Indexing complete.  Post-processing...
[25/Aug/2010:18:34:40 +0200] - import userRoot: Flushing caches...
[25/Aug/2010:18:34:40 +0200] - import userRoot: Closing files...
[25/Aug/2010:18:34:40 +0200] - import userRoot: Import complete.  Processed 9523 entries
in 7 seconds. (1360.43 entries/sec)
[25/Aug/2010:18:34:40 +0200] NSMMReplicationPlugin - multimaster_be_state_change: replica
dc=id,dc=polytechnique,dc=edu is coming online; enabling replication

But  when i continue and try to make a change on a master the consumer
server  crashes.  So the total replica initialisation is ok but even a
single  incremental  update  crashes the consumer server. And there is
nothing  helpful  in logs. I haven't tried the 1.2.6.rc7 version, i've
tried  the latest code version (as of today). Don't know if it matters
(there  seem  to  be  a  lot  of coverity defects that have been fixed
between rc7 and a1).


Andrey Ivanov
tel [phone]
fax [phone]

Direction des Systemes d'Information
Ecole Polytechnique
91128 Palaiseau CEDEX
France

[Rich Megginson]

Andrey Ivanov wrote:
>  I wanted to configure the replication over SSL (both with SSL
>  mechanism which was available in previous versions) and by TLS using
>  simple bind (both in multimaster or single master-dedicated consumer models).
>
>  I've tried to configure it with command line and with the console. The
>  configuration and the initial initialisation are ok :
>
>  [25/Aug/2010:18:30:44 +0200] NSMMReplicationPlugin - replica_config_delete: Warning: The
> changelog for replica dc=id,dc=polytechnique,dc=edu is no longer valid since the replica
> config is being deleted.  Removing the changelog.
>  [25/Aug/2010:18:34:33 +0200] NSMMReplicationPlugin - multimaster_be_state_change: replica
> dc=id,dc=polytechnique,dc=edu is going offline; disabling replication
>  [25/Aug/2010:18:34:33 +0200] - WARNING: Import is running with
> nsslapd-db-private-import-mem on; No other process is allowed to access the database
>  [25/Aug/2010:18:34:39 +0200] - import userRoot: Workers finished; cleaning up...
>  [25/Aug/2010:18:34:40 +0200] - import userRoot: Workers cleaned up.
>  [25/Aug/2010:18:34:40 +0200] - import userRoot: Indexing complete.  Post-processing...
>  [25/Aug/2010:18:34:40 +0200] - import userRoot: Flushing caches...
>  [25/Aug/2010:18:34:40 +0200] - import userRoot: Closing files...
>  [25/Aug/2010:18:34:40 +0200] - import userRoot: Import complete.  Processed 9523 entries
> in 7 seconds. (1360.43 entries/sec)
>  [25/Aug/2010:18:34:40 +0200] NSMMReplicationPlugin - multimaster_be_state_change: replica
> dc=id,dc=polytechnique,dc=edu is coming online; enabling replication
>
>  But  when i continue and try to make a change on a master the consumer
>  server  crashes.  So the total replica initialisation is ok but even a
>  single  incremental  update  crashes the consumer server. And there is
>  nothing  helpful  in logs. I haven't tried the 1.2.6.rc7 version, i've
>  tried  the latest code version (as of today). Don't know if it matters
>  (there  seem  to  be  a  lot  of coverity defects that have been fixed
>  between rc7 and a1).
>     
Can you get a core file and a stack trace?
>  Andrey Ivanov
>  tel [phone]
>  fax [phone]
>
>  Direction des Systemes d'Information
>  Ecole Polytechnique
>  91128 Palaiseau CEDEX
>  France
>
>  --
>  389 users mailing list
>  389-users(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/389-users
>

[Ivanov Andrey (M.)]

2010/8/25 Rich Megginson <rmeggins(a)redhat.com&gt;

>  Andrey Ivanov wrote:
>
> > I wanted to configure the replication over SSL (both with SSL
> > mechanism which was available in previous versions) and by TLS using
> > simple bind (both in multimaster or single master-dedicated consumer
> > models).
> >
> > I've tried to configure it with command line and with the console. The
> > configuration and the initial initialisation are ok :
> >
> > [25/Aug/2010:18:30:44 +0200] NSMMReplicationPlugin -
> > replica_config_delete: Warning: The changelog for replica
> > dc=id,dc=polytechnique,dc=edu is no longer valid since the replica config is
> > being deleted.  Removing the changelog.
> > [25/Aug/2010:18:34:33 +0200] NSMMReplicationPlugin -
> > multimaster_be_state_change: replica dc=id,dc=polytechnique,dc=edu is going
> > offline; disabling replication
> > [25/Aug/2010:18:34:33 +0200] - WARNING: Import is running with
> > nsslapd-db-private-import-mem on; No other process is allowed to access the
> > database
> > [25/Aug/2010:18:34:39 +0200] - import userRoot: Workers finished; cleaning
> > up...
> > [25/Aug/2010:18:34:40 +0200] - import userRoot: Workers cleaned up.
> > [25/Aug/2010:18:34:40 +0200] - import userRoot: Indexing complete.
> >  Post-processing...
> > [25/Aug/2010:18:34:40 +0200] - import userRoot: Flushing caches...
> > [25/Aug/2010:18:34:40 +0200] - import userRoot: Closing files...
> > [25/Aug/2010:18:34:40 +0200] - import userRoot: Import complete.
> >  Processed 9523 entries in 7 seconds. (1360.43 entries/sec)
> > [25/Aug/2010:18:34:40 +0200] NSMMReplicationPlugin -
> > multimaster_be_state_change: replica dc=id,dc=polytechnique,dc=edu is coming
> > online; enabling replication
> >
> > But  when i continue and try to make a change on a master the consumer
> > server  crashes.  So the total replica initialisation is ok but even a
> > single  incremental  update  crashes the consumer server. And there is
> > nothing  helpful  in logs. I haven't tried the 1.2.6.rc7 version, i've
> > tried  the latest code version (as of today). Don't know if it matters
> > (there  seem  to  be  a  lot  of coverity defects that have been fixed
> > between rc7 and a1).
> >
> >
>  Can you get a core file and a stack trace? 

Rich, just as i thought, this crash happens only with today's snapshot of
1.2.7.a1 version only. I've compiled 1.2.6.rc7 and the replication works
smoothly and without any problem. I didn't have a lot of time to generate a
stack trace because i was migrating our production servers. I thought the
latest build should be stable but it seems that the changes between 6rc7 and
7a1 introduce some problems with incremental replication as well as with
apostrophs in DN (my second mail). So for now i will migrate to 1.2.6.rc7.
I'll test the a1 version later when i will have time...

Thanks!

[Rich Megginson]

Andrey Ivanov wrote:
>  2010/8/25 Rich Megginson <rmeggins(a)redhat.com 
>  <mailto:[e-mail]>>
>
>      Andrey Ivanov wrote:
>
>          I wanted to configure the replication over SSL (both with SSL
>          mechanism which was available in previous versions) and by TLS
>          using
>          simple bind (both in multimaster or single master-dedicated
>          consumer models).
>
>          I've tried to configure it with command line and with the
>          console. The
>          configuration and the initial initialisation are ok :
>
>          [25/Aug/2010:18:30:44 +0200] NSMMReplicationPlugin -
>          replica_config_delete: Warning: The changelog for replica
>          dc=id,dc=polytechnique,dc=edu is no longer valid since the
>          replica config is being deleted.  Removing the changelog.
>          [25/Aug/2010:18:34:33 +0200] NSMMReplicationPlugin -
>          multimaster_be_state_change: replica
>          dc=id,dc=polytechnique,dc=edu is going offline; disabling
>          replication
>          [25/Aug/2010:18:34:33 +0200] - WARNING: Import is running with
>          nsslapd-db-private-import-mem on; No other process is allowed
>          to access the database
>          [25/Aug/2010:18:34:39 +0200] - import userRoot: Workers
>          finished; cleaning up...
>          [25/Aug/2010:18:34:40 +0200] - import userRoot: Workers
>          cleaned up.
>          [25/Aug/2010:18:34:40 +0200] - import userRoot: Indexing
>          complete.  Post-processing...
>          [25/Aug/2010:18:34:40 +0200] - import userRoot: Flushing caches...
>          [25/Aug/2010:18:34:40 +0200] - import userRoot: Closing files...
>          [25/Aug/2010:18:34:40 +0200] - import userRoot: Import
>          complete.  Processed 9523 entries in 7 seconds. (1360.43
>          entries/sec)
>          [25/Aug/2010:18:34:40 +0200] NSMMReplicationPlugin -
>          multimaster_be_state_change: replica
>          dc=id,dc=polytechnique,dc=edu is coming online; enabling
>          replication
>
>          But  when i continue and try to make a change on a master the
>          consumer
>          server  crashes.  So the total replica initialisation is ok
>          but even a
>          single  incremental  update  crashes the consumer server. And
>          there is
>          nothing  helpful  in logs. I haven't tried the 1.2.6.rc7
>          version, i've
>          tried  the latest code version (as of today). Don't know if it
>          matters
>          (there  seem  to  be  a  lot  of coverity defects that have
>          been fixed
>          between rc7 and a1).
>           
>
>      Can you get a core file and a stack trace?
>
>   
>  Rich, just as i thought, this crash happens only with today's snapshot 
>  of 1.2.7.a1 version only. I've compiled 1.2.6.rc7 and the replication 
>  works smoothly and without any problem. I didn't have a lot of time to 
>  generate a stack trace because i was migrating our production servers. 
>  I thought the latest build should be stable but it seems that the 
>  changes between 6rc7 and 7a1 introduce some problems with incremental 
>  replication as well as with apostrophs in DN (my second mail). So for 
>  now i will migrate to 1.2.6.rc7. I'll test the a1 version later when i 
>  will have time... 
Thanks for the confirmation.  Yeah, the 1.2.7 (master) is
unstable right 
now.  There have been a lot of changes going in that haven't been fully 
tested with our test suite (we're also having to update the test suite).
>  Thanks!
>