Hello,

We're busy with an initial replication of a large DB to multiple slaves. 
The first slave has completed (after more than a week), but the others are
having what I imagine is a hiccup.

They're repeatedly restarting and copying the same tables - and never
completing.

The relevant section (during a funny restart) from the slony log (on a
slave):

>>> here's the first sign of something amiss (?):
...
01:53:16 FATAL localListenThread: cannot start transaction -
...
01:53:16 DEBUG2 slon_retry() from pid=7750
01:53:17 DEBUG1 slon: retry requested
01:53:17 DEBUG2 slon: notify worker process to shutdown
01:53:17 INFO remoteListenThread_4: disconnecting from 'dbname=xxx
host=zzz3 user=hhh password=000 port=5431'
01:53:17 INFO remoteListenThread_2: disconnecting from 'dbname=xxx
host=zzz1 user=hhh password=000 port=5431'
01:53:17 DEBUG1 syncThread: thread done
01:53:17 INFO remoteListenThread_1: disconnecting from 'dbname=xxx
host=zzz0 user=hhh password=000'
01:53:17 INFO remoteListenThread_3: disconnecting from 'dbname=xxx
host=zzz2 user=hhh password=000 port=5431'
01:53:18 DEBUG1 main: scheduler mainloop returned
01:53:18 DEBUG2 main: wait for remote threads
01:53:18 DEBUG2 sched_wakeup_node(): no_id=1 (0 threads + worker signaled)
01:53:20 DEBUG1 remoteListenThread_3: thread done
01:53:21 DEBUG1 remoteListenThread_1: thread done
01:53:21 DEBUG1 remoteListenThread_2: thread done
01:53:21 DEBUG1 remoteListenThread_4: thread done
01:53:37 DEBUG1 slon: child termination timeout - kill child
01:53:37 DEBUG2 slon: child terminated status: 9; pid: 7750, current
worker pid: 7750
01:53:37 DEBUG1 slon: restart of worker in 10 seconds
...
>>> slony starts again, and proceeds to copy the same tables...
...
01:53:47 CONFIG  main: slon version 1.2.12 starting up
01:53:47 DEBUG2 slon: watchdog process started
01:53:47 DEBUG2 slon: watchdog ready - pid = 7749
01:53:47 DEBUG2 slon: worker process created - pid = 19510
01:53:49 CONFIG  main: local node id = 5
01:53:49 DEBUG2 main: main process started
...

I've also checked the PG logs for any obvious errors.

Any idea where else to look and what to look for?

Any comments welcome.

Thanks
Henry