AWS RDS Postgres backup is using temp file on access node

Hi Gergely,

You need to have sufficient space to stage the dump output file , in windows access node, in both cases. 

Commvault backups, by default, perform pg_dump in windows client using tar format (-Ft). This doesn’t require staging from commvault side but postgresql application use the system temp directory as a staging folder for temporary files when generating tar-based dump files. (typically C:\temp folder)

If commvault service pack level is 28 or higher, to use an alternate folder with sufficient space for these temporary files create additional setting sPgDumpTempDirectory [under PostGres category] to the full path of the desired directory. 

With sPGCompDump set, it dumps database in a compressed output format and the dump is staged at the job results folder by default. This format requires staging the dump output before moving to media agent. To change the staging location create additional setting sPGStagingDir [under Postgres] and setting its value as the new staging location.

Regards,

Meera

Hi Meera,

Thanks for the explanation, that matches the behaviour I have observed, with -Ft it uses C:\Windows\Temp. Is it any different with Linux access node if I use -Fc, does it ‘pipe’ properly without using the cvjobresults folder?

I would prefer -Ft to be honest as compressed dump won’t deduplicate really and I have many TBs to protect daily. Is it possible to redirect the temp directory pg_dump uses with -Ft on Linux?On Linux I could use an EFS mount for this (so I don’t need to provision a big EBS volume, it’s costly to have it around 24/7)

I need to do MySQL as well, what’s the situation with that regarding uncompressed dumps and temp space usage?

Gergely