I'm upgrading a PostgreSQL 9.5 cluster to 9.6 using pg_upgradecluster. It's taking an awful lot of time, so I'd like to cancel the process, delete unneeded databases, and start again. What are the risks?
Add a comment
|
1 Answer
By default, pg_upgradecluster doesn't use pg_upgrade. It's creating a logical dump with pg_dump and restoring it (with pg_restore) on the new cluster.
To avoid this and speed up the process, you need to add the -m upgrade option.
If you need to speed up the process once more, you can add the --link so that the upgarde will be inplace. Be carefull, you can't cancel the process if you did it or you will loose your cluster.
If you didn't use the -m upgrade option and the --link option, you can cancel the process, drop the new cluster and create it again:
pg_dropcluster 9.6 <cluster_name>
pg_createcluster 9.6 <cluster_name>
-
2Thanks very much. I eventually did cancel the process (not using
-m upgradeor--link, thankfully) and found one gotcha: it totally hosed thepg_hba.conffor both the old and new server (I guess it was locking things down for the duration of the upgrade?) It backs it up topg_hba.conf.pg_upgradeclusterthough. May 15, 2017 at 18:17