Re: Backup throttling

On 8/27/13 7:58 AM, Robert Haas wrote:
> We have a *general* need to be able to throttle server-side resource
> utilization, particularly I/O. This is a problem not only for
> pg_basebackup, but for COPY, CLUSTER, VACUUM, and even things like
> UPDATE. Of all of those, the only one for which we currently have any
> kind of a solution is VACUUM.

I didn't mention it specifically, but I always presumed that the "Cost
limited statements RFC" proposal I floated:
http://www.postgresql.org/message-id/519EA5FF.5040606@2ndQuadrant.com
(and am still working on) would handle the base backup case too.
pg_basebackup is just another client. Some sort of purpose made
solution for pg_basebackup alone may be useful, but I'd be shocked if
the sort of general mechanism I described there wasn't good enough to
handle many of the backup limiting cases too.

Also, once that and the block write counters I also sent an RFC out for
are in place, I have a plan for adding a server-wide throttling
mechanism. I want to extract a cluster wide read and write rate and put
a cluster wide limit on that whole thing. It seemed too hard to jump
into without these other two pieces of plumbing in place first.

--
Greg Smith 2ndQuadrant US greg(at)2ndQuadrant(dot)com Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com