Re: Additional DTrace Probes

ITAGAKI Takahiro píše v pá 06. 03. 2009 v 17:03 +0900:
> Hi,
>
> Robert Lor <Robert(dot)Lor(at)Sun(dot)COM> wrote:
>
> > The attached patch contains the probes (originally came from Theo
> > Schlossnagle) that were removed from the previous submitted patch.
> > Zdenek had some concerns about the way the probes were implemented
> > http://archives.postgresql.org/pgsql-hackers/2008-07/msg01168.php. If
> > there are specific recommendations, I'd be more than happy to make the
> > changes and get them resubmitted.
>
> This will introduce SLRU and Executor probes.
> We already have Lock, LWLock, Buffer, I/O and XLogs probes.
>
> I'd like to have the following probes, too. In my experience,
> they could be bottlenecks or consume much time in some situations.

Of course, there are lot of places where probes can be added and it
needs to determine correct place regrading to data which we want to
collect.

> - idle in transaction

This probes lost in a action. I remember that I had some comments about
them. It requires small code refactoring (some extra if or something
like this).

> - spinlock wait-and-retry

Can you specify a place in code or what you need to measure?

> - CPU: Trigger execution
> - CPU: Encoding conversion

Maybe function call start and done could cover both cases.

> - Network: send() and recv()

I think there could be more probes on network like login/logout, login
handshake itself and so on.

> - smgr: lseek() calls

Is any reason for it? You can monitoring syscall directly.

> - Tempfile reads and writes

I think that it is possible now. last arg in buffer_read function
contains information, that it is local(temp) buffer.

Zdenek