From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | "Andrew Hammond" <andrew(dot)george(dot)hammond(at)gmail(dot)com> |
Cc: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: further meta-data in pg_stat_activity? |
Date: | 2006-06-17 04:05:31 |
Message-ID: | 10885.1150517131@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
"Andrew Hammond" <andrew(dot)george(dot)hammond(at)gmail(dot)com> writes:
> How much work would it be to implement and how valuable would people
> find the following additions to pg_stat_activity?
I won't speak to the "how valuable" bit, but as far as costs go,
I think:
> 1) # of transactions committed on this connection since start
> 2) # of transactions rolled back
Trivial, we report these to the stats collector already, they're just
not summed in this particular fashion.
> 3) milliseconds used processing requests
> 4) milliseconds idle in transaction
> 5) milliseconds idle
All moderately expensive, we're talking at least two additional kernel
calls per request to get the information.
> 6) this is the n'th backend spawned since the postmaster started
Cheap on Unix, not so cheap on Windows, usefulness pretty questionable.
> 7) this is the n'th backend for the given client_addr
> 8) this is the n'th backend for the given user
Both *exceedingly* expensive --- where are you going to sum these?
The postmaster does not even have a way to count the second, because
it forks off the subprocess before receiving the connection request
packet which contains the user name.
> 9) timestamp for start of the current transaction (null if idle?)
Don't we do that already?
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Robert Lor | 2006-06-17 04:17:04 | Re: Sun Donated a Sun Fire T2000 to the PostgreSQL community |
Previous Message | Tom Lane | 2006-06-17 02:58:05 | Exporting type OID macros in a cleaner fashion |