Re: Cluster "stuck" in "not accepting commands to avoid wraparound data loss"

On Thu, Dec 10, 2015 at 1:55 PM, Andres Freund <andres@anarazel.de> wrote:

>
> Looking at datfrozenxid:
> postgres=# select datname, datfrozenxid, age(datfrozenxid) FROM pg_database ;
>   datname  | datfrozenxid |    age
> -----------+--------------+-----------
>  template1 |   3357685367 |         0
>  template0 |   3357685367 |         0
>  postgres  |   3159867733 | 197817634
> (3 rows)
> reveals that the launcher doesn't do squat because it doesn't think it
> needs to do anything.
>
> (gdb) p *ShmemVariableCache
> $3 = {nextOid = 24576, oidCount = 0, nextXid = 3357685367, oldestXid = 1211201715, xidVacLimit = 1411201715,
xidWarnLimit= 3347685362,
 
>   xidStopLimit = 3357685362, xidWrapLimit = 3358685362, oldestXidDB = 12380, oldestCommitTs = 0, newestCommitTs = 0,
>   latestCompletedXid = 3357685366}

Do we know how template0 and template1 get frozen with xid which were
5 past the xidStopLimit?  Is that part of the mystery here, or is that
normal?

Cheers,

Jeff