Replication: Log Reader Agent Fails

[eichmat]

Situation:
SERVER-1 (SQL 2000 [no SPs]): Distributor and Subscriber (in US)
SERVER-2 (SQL 2000 [no SPs]): Publisher (in US)
SERVER-3 (SQL 2000 SP 1) : Publisher (in UK)
SERVER-3 (SQL 2000 [no SPs]): Publisher (in CH)

We are using Transactional Replication with continuous update from two overseas servers and one domestic server (located in the same room as SERVER-1). Two databases are pulled from each server.

Problem:
Log Reader Agents for the to overseas servers generate the following error message:

The step did not generate any output. The step failed.

If there aren't any transaction to replicate, the agents run for about 30-seconds before they generate this error. If there were queued transactions, the transactions are replicated, then about 15-seconds later, the agents fail with the above message.

Work-around has been to add a schedule to the Log Reader Agent jobs that "kick-starts" the agent every minute. It works, giving our data up to a 1-minute latency, but... Each time the four log reader agents fail, they dump a message into the Application Log causing it to fill rapidly (~5,700 entries per day).

I tried the microsoft.public.sqlserver.replication newsgroup and all I get is "set verbose logging and post the results", but never a review of the logs with a suggestion or response.

Thanks,


-Tim

 

[MikeTom]

The log reader agent should run as auto start. their are then options to configure the number of times as a retry and a latency time. set retries to 1000 and latency to 30 secs. this means if no output is generated it will retry in 30secs instead of failing with 'The step did not generate any output. The step failed' message. if it retries a 1000 times and still no tasks to replicate (very unlikely) it will then fail.
Their is also an option what to do in an evnt of a failure. if you remove the option to write to nt event logs on success or failure you will stop getting events in event log.