Re: Recommendation to run vacuum FULL in parallel

On 4/3/19 12:50 AM, Perumal Raj wrote:
> Hi ALL
>
> We are  planning to reclaim unused space from 9.2 Version postgres Cluster,
>
> Method : VACUUM FULL

Does *every* table have *so much* free space that it's impractical to just
let the files just get refilled by normal usage?

> DB Size : 500 GB
> Expected space to reclaim 150 GB
> work_mem : 250 MB
> maintenance_work_mem : 20 GB
>
> *Question :*
>
> 1. vacuumdb --j option (Parallel) not available for version 9.2.
>   How to run vacuum full in parallel ? At present its taking 8Hrs if i run
> sequential ( vacuum full verbose;)
>
> 2. If we run vacuum full, Do we need to run REINDEX/ANALYZE exclusively ?
>
> 3. What is the best way to run VACUUM FULL with less window.

A good way to run *any* task like this in parallel is to generate X lists of
objects, and then process each list in parallel.

--
Angular momentum makes the world go 'round.