Errors 625 (can't connect server) and 637 (all replicas not updated)

[lmoriggi]

Hi,
After trying to delete a server named "HEC1", I have the following situation :
- The server is down and uninstalled.
- The server is out of every replica ring.
- The server is still in the Time synchronization process on all servers left (error 625) !
- Impossible to delete the server using C1 (error 637 = all replica not updated). After deleting all child objects of the server, I moved the server object into another context.
- Impossible to get rid of the obituaries requiring a move of HEC1 : one of the obituary points to the server HEC1 ! Error 625.

HEC1 : NW 5.1 SP7 with eDir 8.7.1.
HEC2 - HEC6 : NW 6.5 SP3 with eDir 8.7.3.
4 partitions (1 root and 3 sub ones at the same level), HEC6 Master.

How to get HEC1 ou of the Time synchromization ? (to get rid of the error 625).

 

[TheLad]

Are your servers pointing to HEC1 for Timesync? Check your TIMESYNC.CFG file on one of your servers just to make sure that HEC1 (or it's IP Address) is not in the configured sources list. Was HEC1 a Primary Timesync server, Secondary etc... ?

-----------------------------------------------------
"It's true, its damn true!"
-----------------------------------------------------

 

[lmoriggi]

Thanks for answering,

HEC1 was not a primary time server.
HEC6 is the primary time server and holds also every Master replica.

 

[marvhuffaker]

Are you in the "Replica and Partition" View when trying to delete the server in ConsoleOne? You can't delete the server in regular view.

625 and the timesync errors tell me that the server is not out of the tree like you suggest.



Marvin Huffaker, MCNE

http://www.redjuju.com

 

[lmoriggi]

HEC1 has no more replica and does not appear in any replica ring (DSREPAIR,..., view replica ring), then HEC1 does not show in Console One in the "Partition and Replica view" for any partition.
But HEC1 is still known by the other servers (DSREPAIR, Advanced options, Servers known).

 

[marvhuffaker]

Ok that's fine.. No replicas, but the server object still exists. In Paritition and Replica view, you are probably still able to see the actual server object.

If the server is gone and never coming back, delete the actual server object in consoleone, from the Paritition and Replica view. (you're not just removing a replica from the server, you're removing the actual server object from the tree).

Marvin Huffaker, MCNE

http://www.redjuju.com

 

[lmoriggi]

No way to delete the server object !
See full and short description beelow :

Objective : Migrate HEC1 from NW5.1SP7 to NW6.5 SP3 (delete HEC1, then create with same name).

Before error :
6 servers and 4 partitions (1 root and 3 sub ones at the same level) :
HEC1 : NW5.1 SP7 eDir 8.7.1, Secondary Time server, 4 partitions R/W
HEC2 : NW6.5 SP3 eDir 8.7.3, Secondary Time server, 4 partitions R/W
HEC3 : NW6.5 SP3 eDir 8.7.3, Secondary Time server, 4 partitions R/W
HEC4 : NW6.5 SP3 eDir 8.7.3, Secondary Time server, 4 partitions Read
HEC5 : NW6.5 SP3 eDir 8.7.3, Secondary Time server, 4 partitions R/W
HEC6 : NW6.5 SP3 eDir 8.7.3, Primary Time server, 4 partitions Master

Actions done :
- C1 : delete the 4 partitions on HEC1.
- HEC1 : power off.
- HEC1 : initialize hard disk.
- C1 : delete child objects of HEC1.
- C1 : move HEC1 to an another context (partition change).

Error :
- 637 when trying to delete the object HEC1 in the destination context.
- 625 when running DSREPAIR, Time synchronization on HEC6.

Conclusion :
- HEC1 is no more part of any replica ring (OK).
- HEC1 is still a server known in HEC2 to HEC6 (DSREPAIR, Advanced options, Servers known) !
- HEC1 (source context) has a sub obituary pending to server HEC1 (itself)(DSBROWSE) !
- No more action possible on partitions (error 637) !
- Re-Time stamp obituaries and run backlink proccess gives 625 error on the destination partition (destination context) !

 

[marvhuffaker]

It sounds like your tree is completely hosed. Pending partition operations can painful. You need to stop monkeying with trying to restamp obituaries and other partition operations until you can get to core NDS problems solved. Your actions are only going to make things worse and you could easily destroy your tree.

Go through an NDS Health Check. Here is a link how:

http://support.novell.com/techcenter/articles/cfa03593.html

Don't just assume it's all good. You will find problems. Don't overlook them. Troubleshoot those errors until you get things processing again.

Here is another good TID to follow if you have replicas that haven't processed completely. I can't stress enough, you need to really work with this TID. Don't just think you've already done it. Spend some time with it and hash it out until you get things figured out.

http://support.novell.com/cgi-bin/search/searchtid.cgi?/10050066.htm

This is not an easy problem and there isn't a quick fix. If you have to, call Novell or someone else qualified to deal with your issues. A few hundred bucks is a lot less than a down network. Ideally, you need to get someone either onsite or with remote access into your network so they can see exactly what is going on.



Marvin Huffaker, MCNE

http://www.redjuju.com

 

[lmoriggi]

In the mean time I've done a Service Request to Novell.
The issue is a dead lock case, that no standard tools can fix.
Thanks for answering.