Kbase P89671: After shutting the Target database down with proshut -byF replication is broken
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
29/04/2009 |
|
Status: Verified
SYMPTOM(s):
After shutting the Target and Source databases down replication is broken
Target and Source databases do not synchronise after restarting replication agent and server
Source database is always proshut BEFORE the Target database
Target database shut down with proshut -byF (5316)
Emergency shutdown initiated... (5316)
The Source database is shut down with proshut -by (2248)
Begin normal shutdown (2248)
On restarting the Replication Agent the followin error messages :
Site Replication Agent agent1 is beginning Startup Synchronization at block 0. (10489)
SYSTEM ERROR: Rollforward Apply AI note, BKUPDCTR=68202668, note updctr=68202661. (1028)
Function dsaAIApplyNote failed in rpNOT_ProcessNoteBlock with error -30000. (10428)
Block processing ended in error at block 1582 in area (ai area number). Site Replication cannot continue. (10495)
Unexpected error -30000 returned to function rpAGT_AgentLoop. (10504)
The Site Replication Agent agent1 is ending. (10506)
After Replication Server fails to synchronize after it has successfully made connections to all configured Agents :
Beginning Site Replication synchronization for the Site Replication Agent agent1. (10508)
The Site Replication Agent agent1 is requesting to be terminated. (10699)
The Site Replication Agent agent1 is being terminated. (10700)
All Site Replication Agents have been terminated or have ended. (10702)
Communications error -168 occurred in function rpNLS_SendInSyncToAgent while sending INSYNC. (10491)
The Site Replication Server will shutdown but the source database will remain active. (10698)
FACT(s) (Environment):
All Supported Operating Systems
Fathom Replication
OpenEdge Replication
CAUSE:
Although proshut -byF on the Target database is not wholly responsible for this behaviour, it is certaintly a contributing factor because issuing a -byF yields unpredictable results due to the 'nature' of -byF (not graceful shutdown)
There are unexpected features, which development are aware of but currently unable to identify the exact circumstances underwhich these occur. The UPDTCTR in the target database should ALWAYS be -1 of the UPDTCTR of the ai note, when these BLOCK COUNTERS get out of sync, which -byF may be a contributing factor, synchronization of the Source and Target databases fail.
FIX:
To initially recover from these circumstances, the Target database will have to be re-made from scratch.
Do not use "proshut -byF" as standard in the Target database shutdown scripts. The only time that the target database needs to be FORCED shutdown, is when the Target AGENT has NOT connected to the Source SERVER in the first place and so it will not allow connections until it has.