log_autovacuum

Gregory Stark wrote:

Could I suggest renaming log_autovacuum to log_autovacuum_min_duration?

Sure, whatever makes the most sense. In fact min_duration would be more
consistent.

--
Alvaro Herrera http://www.amazon.com/gp/registry/CTMLCN8V17R4
"El d�a que dejes de cambiar dejar�s de vivir"

Alvaro Herrera <alvherre@commandprompt.com> writes:

Gregory Stark wrote:

Could I suggest renaming log_autovacuum to log_autovacuum_min_duration?

Sure, whatever makes the most sense. In fact min_duration would be more
consistent.

I'm not sure I believe Greg's argument about needing more autovac
logging parameters, but since this one acts just like
log_min_duration_statement, I concur with renaming it.

regards, tom lane

Actually, we happen to be running into a situation here where we need more
logging. We need to understand why autovacuum isn't considering logging this
table:

relid | schemaname | relname | seq_scan | seq_tup_read | idx_scan | idx_tup_fetch | n_tup_ins | n_tup_upd | n_tup_del | n_live_tup | n_dead_tup | last_vacuum | last_autovacuum | last_analyze | last_autoanalyze
-------+------------+------------+----------+--------------+----------+---------------+-----------+-----------+-----------+------------+------------+-------------+-------------------------------+--------------+-------------------------------
16436 | public | stock | 0 | 0 | 45929274 | 45928278 | 0 | 12116286 | 0 | 25036190 | 12723033 | | | | 2007-08-01 17:24:30.796874-07

It looks like there are some DEBUG3 messages which would be useful but I don't
know of any convenient way to change the log level in autovacuum workers.

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com

On Fri, 2007-08-03 at 12:38 -0400, Tom Lane wrote:

Alvaro Herrera <alvherre@commandprompt.com> writes:

Gregory Stark wrote:

Could I suggest renaming log_autovacuum to log_autovacuum_min_duration?

Sure, whatever makes the most sense. In fact min_duration would be more
consistent.

I'm not sure I believe Greg's argument about needing more autovac
logging parameters, but since this one acts just like
log_min_duration_statement, I concur with renaming it.

log_min_duration_autovacuum

makes the most sense in comparison, IMHO.

--
Simon Riggs
EnterpriseDB http://www.enterprisedb.com

On Fri, 2007-08-03 at 18:56 +0100, Gregory Stark wrote:

Actually, we happen to be running into a situation here where we need more
logging. We need to understand why autovacuum isn't considering logging this
table:

relid | schemaname | relname | seq_scan | seq_tup_read | idx_scan | idx_tup_fetch | n_tup_ins | n_tup_upd | n_tup_del | n_live_tup | n_dead_tup | last_vacuum | last_autovacuum | last_analyze | last_autoanalyze
-------+------------+------------+----------+--------------+----------+---------------+-----------+-----------+-----------+------------+------------+-------------+-------------------------------+--------------+-------------------------------
16436 | public | stock | 0 | 0 | 45929274 | 45928278 | 0 | 12116286 | 0 | 25036190 | 12723033 | | | | 2007-08-01 17:24:30.796874-07

It looks like there are some DEBUG3 messages which would be useful but I don't
know of any convenient way to change the log level in autovacuum workers.

It also appears that only a single autovacuum daemon active at any one
time, which is also weird when we are supposed to have 3.

--
Simon Riggs
EnterpriseDB http://www.enterprisedb.com

On Aug 3, 2007, at 14:59 , Simon Riggs wrote:

On Fri, 2007-08-03 at 12:38 -0400, Tom Lane wrote:

Alvaro Herrera <alvherre@commandprompt.com> writes:

Gregory Stark wrote:

Could I suggest renaming log_autovacuum to
log_autovacuum_min_duration?

Sure, whatever makes the most sense. In fact min_duration would
be more
consistent.

I'm not sure I believe Greg's argument about needing more autovac
logging parameters, but since this one acts just like
log_min_duration_statement, I concur with renaming it.

log_min_duration_autovacuum

makes the most sense in comparison, IMHO.

True, but the log_min_duration_statement is kind of poorly named (as
is log_min_error_statement). log_statement is the overall concept,
min_duration and min_error further specialize the concept.
log_statement_min_duration and log_statement_min_error would have
been better, IMO. Question is whether it's better to move forward
with consistent naming or improve naming when the chance arises.

Michael Glaesemann
grzm seespotcode net