We have a master and a slave DB server (both PostgreSQL 8.3, Slony 2.0.1). Maybe we were hasty, because we delete ~18 million rows from a table on master with one SQL command. Slony log and transfer jobs/transactions to the slave node, but on the slave node the slony can't do this job. Slony starts the transaction, but after ~7 million delete commands the server close the connection.
We found this in the Slony log: 2011-11-15 11:39:25 CET DEBUG4 remoteHelperThread_1_1: fetch from cursor 2011-11-15 11:39:25 CET ERROR remoteHelperThread_1_1: "fetch 500 from LOG; " server closed the connection unexpectedly This probably means the server terminated abnormally before or while processing the request. 2011-11-15 11:39:25 CET DEBUG4 remoteHelperThread_1_1: return 50 unused line buffers 2011-11-15 11:39:25 CET ERROR remoteWorkerThread_1: "close LOG; " PGRES_FATAL_ERROR 2011-11-15 11:39:25 CET ERROR remoteWorkerThread_1: "rollback transaction; set enable_seqsca n = default; set enable_indexscan = default; " PGRES_FATAL_ERROR 2011-11-15 11:39:25 CET DEBUG1 remoteHelperThread_1_1: 2850.992 seconds until close cursor 2011-11-15 11:39:25 CET INFO remoteHelperThread_1_1: inserts=0 updates=0 deletes=7220000
We tried to tuning the PostgreSQL and now two times faster, but every transaction stop after ~ 7 million rows. We couldn't find error message in the PostgreSQL log.