Merge replication "errors"

Merge replication "errors"

Post by Michael Hote » Sat, 17 Jul 1999 04:00:00

I've seen a lot of messages posted about getting various error messages from
the merge agents.  I've seen all of these, but have keep my mouth shut until
I could investigate them further.  Some of these are as follows: publication
xxx does not exist, number of retries exceed, excessive retries, could not
make a generation, time out, etc.  All of these are non fatal for merge
replication.  I don't have a definitive answer at this point simply because
when I get one of these, I'm usually getting several different ones all at
the same time.  All of these are symptoms of a merge system that is plain
and simply overwhelmed.  The retries and could not make a generation are
most likely due to blocking.  During the merge process a significant amount
of logging is done in the merge tables.  If you have several levels of
merge, you are possibly making changes to the same merge row in those
tables.  This will produce blocking and force a kick out where it attempts
to retry.  Could not make a generation is most likely due to the fact that
it could not modify the generation flag on a set of rows because they were
locked.  The rest have occurred as the system gets more and more
overwhelmed.  The merge agents get lost when you really pressure them and
will generally throw a publication xxx does not exist at which point they
quit processing transactions.

Fixing this is relatively straightforward.  Stop all of the merge agents at
each level in your architecture.  Then simply go to the lowest level and
start one agent back up.  Let it flush all of the transactions through until
it reaches an idle state.  Stop that agent and repeat the process with the
next agent.  When you have cycled through all agents at the lowest level,
move up to the next level and repeat the process.  Continue this process
until you reach the top level.  Once the agents at the top level have
flushed, start all of the top level agents.  Move down to the next level and
start all of those agents.  Continue until you reach the bottom level and
all merge agents are restarted.  Then simply monitor the system.  If you
happen to bottleneck it again, repeat the process until it is all back up
and running.  If you repeatedly do this, you do not have enough power behind
the servers and simply need to throw more hardware at the problem.  The
further up in the hierarchy that you go, the more powerful the server needs
to be.

MCSE, SQL Server DBA, Author

Please direct all replies to the newsgroup unless specifically stated


Merge replication "errors"

Post by Tom Case » Sat, 17 Jul 1999 04:00:00

If you're seeing frequent failures to create generations or excessive
retries on your merge agents, you may need to increase the -QueryTimeout
value in the merge agent profile associated with your agents.


Merge replication "errors"

Post by Michael Hote » Tue, 20 Jul 1999 04:00:00

I'll give that a try.  With the tests I'm running right now, I don't think
it'll matter.  I'm running some stress tests and am seeing a lot of these.
But, I have some very low end hardware and am pounding 2000 transactions a
minute across 6 databases 3 levels down in a hierarchy.  It's taking about
10 minutes to completely overwhelm the system.  But, if I can get it to
sustain even through the retries without hitting an actual failure, then
that is a volume well beyond what we need.



1. Merge Replication "General Network Error"

this is the message that i'm geting

The process could not query row metadata at the 'Subscriber'.
(Source: Merge Replication Provider (Agent); Error number: -2147200996)
General network error. Check your network documentation.
(Source: JETPACKSERVER (Data source); Error number: 11)
The process was successfully stopped.
(Source: Merge Replication Provider (Agent); Error number: -2147199469)

2. Netscape Commerce Server & Sybase SQL server

3. Merge replication error "rowguidcol"

4. Sql*loader error message 523

5. Merge Replication "Q"

6. Copy data with into password protected database

7. Merge Replication "Feature"

8. What should FILEGROWTH be...

9. max of ("...","...","..")

10. Error when configure "Transactional Replication"

11. Snapshot replication gives "Invalid Descriptor Error"

12. Replication error "Unable to apply retries"

13. "not data found" error in replication