Thu Jun 25 04:24:29 PDT 2009
- Previous message: [Slony1-general] Subscription errors don't automatically recover
- Next message: [Slony1-general] Lowering Slony update cpu usage
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hello all, Platform: centOS Linux 4.7 PostgreSQL: 8.3.6 Slony: 1.2.15 Has anyone run into the problem of recovering the failed node after failover process when replication includes large tables? There is the following situation: We have 2 nodes running in master-slave configuration with "hot" backup mode. When master node failes, the slave should become active. The problem is that that we may have large tables in the set and it will take too much time to subscribe the failed node from scratch. Are there any solutions for such kind of situation? May be it is possible to override the node subscription procedure (COPY of all the data in the set) with some kind of self-written procedure which will copy only the most needed (recent) data, and then copy the rest in background and take the responsibility and risks for node sync? The system is kind of critical application, which needs a backup node ready to go. Also, is there any solution to prevent OS from reboot or shutdown until master-slave switchover process is completed? It's very uncomfortable that when we accidentally reboot master machine, our software is receiving soft termination signal at first and tries to do a switchover, but there is a pretty big chance that it won't be completed before all processes (including slony and postgres) receive kill signal and being aborted. Best regards, Nick.
- Previous message: [Slony1-general] Subscription errors don't automatically recover
- Next message: [Slony1-general] Lowering Slony update cpu usage
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Slony1-general mailing list