On Thu, Sep 22, 2005 at 03:52:21PM +1000, Gavin Sherry wrote:
On Wed, 31 Aug 2005, Tom Lane wrote:
BTW ... the original Berkeley papers on Postgres make frequent reference
to a vacuum daemon, which seems to be essentially what we're trying to
build with autovacuum. Does anyone know if
On Sep 22, 2005, at 3:55 PM, David Fetter wrote:
On Thu, Sep 22, 2005 at 03:52:21PM +1000, Gavin Sherry wrote:
On Wed, 31 Aug 2005, Tom Lane wrote:
Well, I was just poking around the executor and noticed this in
ExecDelete():
/*
* Note: Normally one would think that we have to
On Wed, 31 Aug 2005, Tom Lane wrote:
BTW ... the original Berkeley papers on Postgres make frequent reference
to a vacuum daemon, which seems to be essentially what we're trying to
build with autovacuum. Does anyone know if the Berkeley implementation
ever actually had auto vacuuming, or was
On Wed, 2005-08-31 at 22:21 -0400, Tom Lane wrote:
Simon Riggs [EMAIL PROTECTED] writes:
On Wed, 2005-08-31 at 19:24 -0400, Tom Lane wrote:
If you don't remove any tuples,
you don't scan the indexes anyway IIRC.
No. Even if you remove *zero* tuples, an index is still scanned twice.
Josh Berkus josh@agliodbs.com writes:
So, will per-table XID tracking allow us to avoid *ever* vacuuming some
tables?
If your definition of ever is less than a billion transactions, sure.
(As Simon points out, with time-partitioned data sets that could often
be arranged, so it's not a
On Thu, Sep 01, 2005 at 04:21:58AM -, Andrew - Supernews wrote:
On 2005-09-01, Alvaro Herrera [EMAIL PROTECTED] wrote:
On Thu, Sep 01, 2005 at 01:57:02AM +0100, Simon Riggs wrote:
If you're using autovacuum then the problem is already taken care of.
autovacuum will respond only to
On Thu, 2005-09-01 at 10:29 +0100, Simon Riggs wrote:
On Wed, 2005-08-31 at 22:21 -0400, Tom Lane wrote:
Simon Riggs [EMAIL PROTECTED] writes:
On Wed, 2005-08-31 at 19:24 -0400, Tom Lane wrote:
If you don't remove any tuples,
you don't scan the indexes anyway IIRC.
No. Even if
Simon Riggs [EMAIL PROTECTED] writes:
My first proposal is to add an extra parameter onto the
index_bulk_delete() call - ntuples. If ntuples == 0 then btbulkdelete()
will avoid scanning and return immediately. If a scan occurs, then we
keep track of how many tuples have been marked deleted and
For a while now, I've been seeking a method of reducing the impact of
VACUUM when run against a database where 80-95% of the data is
effectively read only and will not be deleted/updated again. This is the
situation in most Data Warehouses. When you get very large databases
(VLDB) the execution
Simon Riggs [EMAIL PROTECTED] writes:
A new command is proposed - VACUUM MINIMAL. The *sole* purpose of this
command is to do the absolute minimum required to avoid transaction id
wraparound. (Better names welcome)
I do not see the point. If you only need to run it every billion
On Wed, 2005-08-31 at 19:24 -0400, Tom Lane wrote:
Simon Riggs [EMAIL PROTECTED] writes:
A new command is proposed - VACUUM MINIMAL. The *sole* purpose of this
command is to do the absolute minimum required to avoid transaction id
wraparound. (Better names welcome)
I do not see the
On Thu, Sep 01, 2005 at 01:57:02AM +0100, Simon Riggs wrote:
If you're using autovacuum then the problem is already taken care of.
autovacuum will respond only to UPDATEs and DELETEs. In the scenario I
outline, these will *never* occur on the largest tables. A VACUUM would
still eventually
Simon Riggs [EMAIL PROTECTED] writes:
On Wed, 2005-08-31 at 19:24 -0400, Tom Lane wrote:
If you don't remove any tuples,
you don't scan the indexes anyway IIRC.
No. Even if you remove *zero* tuples, an index is still scanned twice.
Once to not delete the rows and once to not delete the
I really really do not like proposals to introduce still another kind
of VACUUM. We have too many already; any casual glance through the
archives will show that most PG users don't have a grip on when to use
VACUUM FULL vs VACUUM. Throwing in some more types will make that
problem exponentially
Christopher Kings-Lynne [EMAIL PROTECTED] writes:
I really really do not like proposals to introduce still another kind
of VACUUM. We have too many already; any casual glance through the
archives will show that most PG users don't have a grip on when to use
VACUUM FULL vs VACUUM. Throwing in
On Wed, 31 Aug 2005, Tom Lane wrote:
Christopher Kings-Lynne [EMAIL PROTECTED] writes:
I really really do not like proposals to introduce still another kind
of VACUUM. We have too many already; any casual glance through the
archives will show that most PG users don't have a grip on when
Gavin, Tom,
Well, from my reading of some of the early papers, VACUUM was kind of
different to what it is now. The idea was that expired data would be moved
out the heap and stored else where. A timetravel mechanism could be used
to see previous versions of the row.
And from talking to a
Tom,
If you're using autovacuum then the problem is already taken care of.
It will be taken care of better in 8.2, if we add per-table tracking
of XID wraparound risk, but it's handled now. The only way that this
proposal makes any sense is if you are trying not to vacuum at all, ever.
Hmmm
On 2005-09-01, Alvaro Herrera [EMAIL PROTECTED] wrote:
On Thu, Sep 01, 2005 at 01:57:02AM +0100, Simon Riggs wrote:
If you're using autovacuum then the problem is already taken care of.
autovacuum will respond only to UPDATEs and DELETEs. In the scenario I
outline, these will *never* occur
19 matches
Mail list logo