Poor Replication Server performance can sometimes be caused by
the Stable Queue Transaction (SQT) cache size being too small.
If
the SQT cache is not big enough to hold all open transactions, transactions
that cannot completely fit into the cache are processed one command
at a time. The inbound stable queues grow (more than 1MB) because Replication
Server cannot process the transactions fast enough.
Solution
Increase the stable queue size to process more transactions and improve Replication Server performance:
- Make
sure that the problem is not caused by orphaned transactions.
- Find the optimal size for the sqt_max_cache_size parameter
for your application. See Replication Server Administration
Guide Volume 2.
- Suspend the connection to the
data server.
- Log in to Replication Server and run:
> configure replication server
set sqt_max_cache_size to ‘new_value’
> go
- To activate the new values, resume the connection.
If you want to dump the SQT cache, use the sysadmin
sqt_dump_queue command.