problems with merge replication

I have big problems with a merge replication. The replication stops very
often with different error messages. The error messages cannot be reproduced
and sometimes the replication finishes without any problem or stops at
another step. Because of that I think the reason must be something different
and cannot be found in the error messages (But maybe I am wrong).
First I thought it could be some sort of timeout but to change the login and
query timeouts did not change anything.
What can be the reason for that? Are there any internal timeouts, that
cannot be configured by the user? Could it be the row filters or some
configuration mismatch?
Another Problem is the following: Applying a snapshot (even over 56k modem )
normally takes less than 5 minutes. But sometimes it takes more than 90
minutes, because all articles are processed several times ( up to 40 times
and more!).
Example output (translated from the german version, I don’t know how the
corresponding output in the english version is):
Processing article ‘A’ …
Processing article ‘B’ …
Processing article ‘C’ …
Processing article ‘A’ …
Processing article ‘B’ …
Processing article ‘C’ …
Processing article ‘A’ …
Processing article ‘B’ …
Processing article ‘C’ …
…..
Why are the articles processed multiple times – one time for each article
should be enough? What makes the merge agent loop through all the articles
again and again? Maybe someone could explain what the merge agent
(replmerg.exe) internally does and checks? Why does it stop after X loops?
Following some additional information:
There is a central data base in the indoor service. About 40 field
representatives are to make a data alignment with laptops.
In the indoor service the SQL server runs 2000 SP3 on a Windows of 2000
servers with all current Microsoft updates for Windows of 2000 servers and
SQL servers 2000. On the laptops runs the MSDE 2000 SP3 on Windows 2000.
Apart from the computer name all laptops are equivalent exactly configured
and equipped.
The merge replication has row filters and for some tables also column
filters. For each laptop a dynamic snapshot is produced. For the replication
pull subscriptions are used. Each laptop replicates at least once per week.
Before the replication is started, I transfer the files for the dynamic
Snapshot to the laptop. To start the replication I use the replmerg.exe of
the msde installation on the laptop. The laptops are connected either by the
LAN or by phone line. The problem with the abortion of the replication seems
to occur more frequently if the bandwidth of the connection is lower (over
the LAN it mostly works fine). The replication with the modem is more
important, because most field representatives are external.
Who had already once such problems or can help me with these problems? If
you need more detailed information please tell me what you need.
Thomas Boll
ESTEAM Software GmbH
http://www.esteam.de
Many merge replication agent failures can be solved simply by running the
agent again. These errors are often transitory.
When you get the processing a, processing b, processing a, it means that
there has been an error with applying a child record before a parent or a
similar dri related conflict. Use conflict viewer to see if these conflicts
have been recorded there.
If they have not, the conflict was probably resolved when the agent retried
the commands.
Lets go over an example
Merge agent starts
Processes article 1, inserts a child record with no matching parent, command
goes on queue to be tried again
Processes article 2, inserts the missing parent
Processes article 3, everything works
Merge agents finishes processing all articles, goes to the retry queue
Processes article 1, inserts the child record which bombed the first time,
this time it succeeds.
Merge agent completes
HTH
Hilary Cotter
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html
"Thomas Boll" <b0ll@.esteam.de_replace0withO.de> wrote in message
news:6B40110B-BEE2-4A90-A749-1092E34C2185@.microsoft.com...
> I have big problems with a merge replication. The replication stops very
> often with different error messages. The error messages cannot be
reproduced
> and sometimes the replication finishes without any problem or stops at
> another step. Because of that I think the reason must be something
different
> and cannot be found in the error messages (But maybe I am wrong).
> First I thought it could be some sort of timeout but to change the login
and
> query timeouts did not change anything.
> What can be the reason for that? Are there any internal timeouts, that
> cannot be configured by the user? Could it be the row filters or some
> configuration mismatch?
> Another Problem is the following: Applying a snapshot (even over 56k
modem )
> normally takes less than 5 minutes. But sometimes it takes more than 90
> minutes, because all articles are processed several times ( up to 40 times
> and more!).
> Example output (translated from the german version, I don't know how the
> corresponding output in the english version is):
> Processing article 'A' .
> Processing article 'B' .
> Processing article 'C' .
> Processing article 'A' .
> Processing article 'B' .
> Processing article 'C' .
> Processing article 'A' .
> Processing article 'B' .
> Processing article 'C' .
> ...
> Why are the articles processed multiple times - one time for each article
> should be enough? What makes the merge agent loop through all the articles
> again and again? Maybe someone could explain what the merge agent
> (replmerg.exe) internally does and checks? Why does it stop after X loops?
> Following some additional information:
> There is a central data base in the indoor service. About 40 field
> representatives are to make a data alignment with laptops.
> In the indoor service the SQL server runs 2000 SP3 on a Windows of 2000
> servers with all current Microsoft updates for Windows of 2000 servers and
> SQL servers 2000. On the laptops runs the MSDE 2000 SP3 on Windows 2000.
> Apart from the computer name all laptops are equivalent exactly configured
> and equipped.
> The merge replication has row filters and for some tables also column
> filters. For each laptop a dynamic snapshot is produced. For the
replication
> pull subscriptions are used. Each laptop replicates at least once per
week.
> Before the replication is started, I transfer the files for the dynamic
> Snapshot to the laptop. To start the replication I use the replmerg.exe of
> the msde installation on the laptop. The laptops are connected either by
the
> LAN or by phone line. The problem with the abortion of the replication
seems
> to occur more frequently if the bandwidth of the connection is lower (over
> the LAN it mostly works fine). The replication with the modem is more
> important, because most field representatives are external.
> Who had already once such problems or can help me with these problems? If
> you need more detailed information please tell me what you need.
> Thomas Boll
> --
> ESTEAM Software GmbH
> http://www.esteam.de