My specific case:
I updated every row in a large table.
In the same transaction I tried to run vacuum on this table, which
rolled back the transaction. Then I reran my update. So this large
table is now 2/3 expired tuples, and the performance hit is noticable.
Bill Moran wrote:
> Joseph Shraibman wrote:
>
>> If I cancel a VACUUM FULL, is the work that was done up until that
>> point thrown away? I have a table that needs vacuuming but I can't
>> accept the downtime involved in vacuuming.
>
>
> Not sure about the "cancel vacuum full" question, but I had some other
> thoughts
> for you.
>
> Keep in mind that a plain vacuum can do a lot of good if done regularly,
> and
> it doesn't lock tables, thus the database can be in regular use while it's
> run. As a result, there is no downtime involved with regularly scheduled
> vacuums.
>
> There _can_ be a performance hit while vacuum is running, so you may
> need to
> take that into account. But I would expect that the performance hit
> incurred
> during running vacuum will be less than that of not running it for long
> periods of time.
>
