On 2/4/16 9:46 AM, Tore Halvorsen wrote: > On Thu, Feb 4, 2016 at 3:30 PM, Scott R Mead <scottm@xxxxxxxxxxx > <mailto:scottm@xxxxxxxxxxx>> wrote: > >> > Can a hot standby run correctly with the xlog-files on a >> ramdisk? >> > >> >> Yes, but, if you lose any, you'll need to rebuild the standby. >> >> >> Even if the master and archive contain all the wal files? >> > If you have the file somewhere, you're okay, whether it's on the > master, standby or an archive host. If you're using log-shipping, > you'll need all the WALs. > > > So, just to verify my understanding. Using a ramdisk for pg_xlog on a > hot standby slave should be ok in all cases as long as the archive and > master retains all wal files that are not applied? Yes. If the pg_xlog ramdisk is lost on the standby then Postgres will start throwing errors but there will be no corruption of the heap. It's possible that Postgres will terminate at this point, but after remounting the ramdisk you can restart Postgres on the standby and everything will be fine. If something goes really horribly wrong you can always rebuild the standby. -- -David david@xxxxxxxxxxxxx
Attachment:
signature.asc
Description: OpenPGP digital signature
