On 2/6/07, Alvaro Herrera <alvherre@xxxxxxxxxxxxxxxxx> wrote:
Merlin Moncure wrote:
> On 2/6/07, Alvaro Herrera <alvherre@xxxxxxxxxxxxxxxxx> wrote:
> >> actually, here is some more relevant bits from the log.
> >> Feb 6 06:31:33 mojo postgres[1088]: [1-1] :: LOG: autovacuum:
> >> processing database "template0"
> >> Feb 6 06:31:33 mojo postgres[1088]: [2-1] :: ERROR: could not access
> >> status of transaction 544441911
> >> Feb 6 06:31:33 mojo postgres[1088]: [2-2] :: DETAIL: could not open
> >> file "pg_clog/0207": No such file or directory
> >
> Latest checkpoint's NextXID: 2162841139
> 2^31: 2147483648
I think the relevant arithmetic here is
echo "2162841139 544441911 - p" | dc
1618399228
That's a billion and a half transactions. Autovacuum uses the formula
this_whole_db = (tmp->age >
(int32) ((MaxTransactionId >> 3) * 3 - 100000));
to determine whether it needs database-wide vacuum.
(MaxTransactionId >> 3) is 536870911, so the calculation is
536870911 * 3 - 100000
echo "536870911 3 * 100000 - p" | dc
1610512733
which looks awfully close to the number above. About 7 million
transactions must have passed since the first time the error showed up
-- does that sound likely?
Well, scratch that -- what's the _current_ Xid? (not lastest
checkpoint's)
I don't know any better way to get that than this:
postgres=# insert into foo default values;
INSERT 0 1
postgres=# select xmin,xmax from foo;
xmin | xmax
------------+------
2163877346 | 0
(1 row)
merlin
