Whenever I read the documentation on the transaction wraparound stuff,
my head spins with all the references to frozen xids and min/max ages.
When it comes down to it, my interpretation of the whole kaboodle is
...."run postgresql v8.1 or later and autovacuum, and you will not have
to deal with the wraparound issue."
I was looking at some logs this morning and spotted this on my backup
database
Jun 24 04:08:46 back-app-1 postgres[82445]: [2-1] pgsql 82445 LOG: transaction ID wrap limit is 41612954, limited by
database"back"
That seemed like an awfully low number compared to the 1 or 2 billion that I have seen in other posts on the subject.
postgres=# SELECT datname, age(datfrozenxid) FROM pg_database;
datname | age
-----------+------------
postgres | 1073968236
back | 1079399370
template0 | 1657840460
template1 | 1078788693
postgres=# select * from pg_database;
datname | datdba | encoding | datistemplate | datallowconn | datconnlimit | datlastsysoid | datvacuumxid |
datfrozenxid| dattablespace | datconfig | datacl
-----------+--------+----------+---------------+--------------+--------------+---------------+--------------+--------------+---------------+-----------+------------------
postgres | 10 | 0 | f | t | -1 | 10792 | 3268269560 |
2194527737| 1663 | |
back | 10 | 0 | f | t | -1 | 10792 | 3268269655 |
2194527832| 1663 | |
template0 | 10 | 0 | t | f | -1 | 10792 | 1610655513 |
1610655513| 1663 | | {pgsql=CT/pgsql}
template1 | 10 | 0 | t | t | -1 | 10792 | 3268490000 |
2194754451| 1663 | | {pgsql=CT/pgsql}
(4 rows)
Should I be worried?