Tue Sep 18 16:05:52 PDT 2007
- Previous message: [Slony1-general] bug in deadlock handling?
- Next message: [Slony1-general] bug in deadlock handling?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Tue, 18 Sep 2007, Jeff Frost wrote: > Hi guys, I've got an interesting situation on a slony 1.2.10 3 node cluster. > Both slaves get their data direct from the master. Everything has been > running well up to a few days ago. Now every time we try to add a new table > to the cluster, we end up with the following error: > > 2007-09-18 15:56:06 EDT ERROR remoteWorkerThread_1: "select > "_nerdcluster".ddlScript_prepare_int(1, -1); " PGRES_FATAL_ERROR ERROR: > Slony-I: alterTableRestore(): Table "public"."carts" is not in altered state > CONTEXT: SQL statement "SELECT "_nerdcluster".alterTableRestore( $1 )" > PL/pgSQL function "ddlscript_prepare_int" line 46 at perform > > It looks like the problem is being caused by a deadlock: > > 15:27:54 sql1 slon[12252]: [39-1] 2007-09-18 15:27:54 EDT ERROR > remoteWorkerThread_1: "select "_nerdcluster".ddlScript > _complete_int(1, -1); " PGRES_FATAL_ERROR > Sep 18 15:27:54 sql1 slon[12252]: [39-2] ERROR: deadlock detected > Sep 18 15:27:54 sql1 slon[12252]: [39-3] DETAIL: Process 12263 waits for > AccessExclusiveLock on relation 121589880 of databas > e 121589046; blocked by process 12096. > Sep 18 15:27:54 sql1 slon[12252]: [39-4] Process 12096 waits for > AccessShareLock on relation 121589817 of database 121589046; > blocked by process 12263. > > So, my theory is that the execute script alters the tables back to their > normal states, doesn't get all the locks it wants and bails out without > putting them back to their previously altered state, thus breaking > replication. > > So, is there a reasonable way to fix this without droppping/resubscribing the > node? BTW, this is a quiet error...i.e. slonik does not indicate anything untoward happened. Also, the problem only occurs on node 2 and not on node 3. This isn't surprising as there's no activity on node 3 it's just a warm standby, but node 2 is actively processing queries. -- Jeff Frost, Owner <jeff at frostconsultingllc.com> Frost Consulting, LLC http://www.frostconsultingllc.com/ Phone: 650-780-7908 FAX: 650-649-1954
- Previous message: [Slony1-general] bug in deadlock handling?
- Next message: [Slony1-general] bug in deadlock handling?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Slony1-general mailing list