Re: Re: Hot Standby query cancellation and Streaming Replication integration

All,

First, from the nature of the arguments, we need to eventually have both
versions of SR: delay-based and xmin-pub. And it would be fantastic if
Greg Smith and Tom Lane could work on xmin-pub to see if we can get it
ready as well.

I also think, based on the discussion and Greg's test case, that we
could do two things which would make the shortcomings of delay-based SR
a vastly better experience for users:

1) Automated retry of cancelled queries on the slave. I have no idea
how hard this would be to implement, but it makes the difference between
writing lots of exception-handling code for slave connections
(unacceptable) to just slow response times on the slave (acceptable).

2) A more usable vacuum_defer_cleanup_age. If it was feasible for a
user to configure the master to not vacuum records less than, say, 5
minutes dead, then that would again offer the choice to the user of
slightly degraded performance on the master (acceptable) vs. lots of
query cancel (unacceptable). I'm going to test Greg's case with
vacuum_cleanup_age used fairly liberally to see if this approach has merit.

Why do I say that "lots of query cancel" is "unacceptable"? For the
simple reason that one cannot run the same application code against an
HS+SR cluster with lots of query cancel as one runs against a standalone
database. And if that's true, then the main advantage of HS+SR over
Slony and Londiste is gone. MySQL took great pains to make sure that
you could run the same code against replicated MySQL as standalone, and
that was based on having a fairly intimate relationship with their users
(at the time, anyway).

Another thing to keep in mind in these discussions is the
inexpensiveness of servers today. This means that, if slaves have poor
performance, that's OK; one can always spin up more slaves. But if each
slave imposes a large burden on the master, then that limits your
scalability.

--Josh Berkus