Thread: PostgreSQL service won't start after bad computer time

On 6/3/11 8:52:15 AM, Rob Richardson wrote:
> I did some testing involving changing a computer’s time, and left the
> time one year early (6/3/2010 instead of 2011). The PostgreSQL service
> now will not start up. Here’s what the log says:
>
> 2011-06-03 08:46:50 EDTWARNING: autovacuum not started because of
> misconfiguration
>
> 2011-06-03 08:46:50 EDTHINT: Enable the "track_counts" option.
>
> 2011-06-03 08:46:50 EDTLOG: database system shutdown was interrupted;
> last known up at 2011-06-03 08:43:59 EDT
>
> 2011-06-03 08:46:50 EDTLOG: database system was not properly shut down;
> automatic recovery in progress
>
> 2011-06-03 08:46:50 EDTLOG: redo starts at 9/6E08AAB8
>
> 2011-06-03 08:46:50 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:50 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:51 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:52 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:53 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:54 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:55 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:55 EDTFATAL: the database system is starting up
>
> 2011-06-03 08:46:56 EDTLOG: record with zero length at 9/8E080968
>
> 2011-06-03 08:46:56 EDTLOG: redo done at 9/8E07FA78
>
> 2011-06-03 08:46:56 EDTLOG: last completed transaction was at log time
> 2010-06-03 08:11:17.531-04
>
> 2011-06-03 08:46:56 EDTFATAL: xlog flush request 18/A2BE3510 is not
> satisfied --- flushed only to 9/8E080968
>
> 2011-06-03 08:46:56 EDTCONTEXT: writing block 0 of relation global/1261_vm
>
> 2011-06-03 08:46:56 EDTLOG: startup process (PID 4596) exited with exit
> code 1
>
> 2011-06-03 08:46:56 EDTLOG: aborting startup due to startup process failure
>
> How do we recover from this?

Reset the time to be correct.

You can't seriously expect something as time-critical as a RDBMS to
function properly if you jockey around making the system time all
fictional?

--
Bill Moran

Bill Moran <wmoran@potentialtech.com> writes:
> On 6/3/11 8:52:15 AM, Rob Richardson wrote:
>> I did some testing involving changing a computer�s time, and left the
>> time one year early (6/3/2010 instead of 2011). The PostgreSQL service
>> now will not start up. Here�s what the log says:
>>
>> 2011-06-03 08:46:56 EDTLOG: record with zero length at 9/8E080968
>> 2011-06-03 08:46:56 EDTLOG: redo done at 9/8E07FA78
>> 2011-06-03 08:46:56 EDTLOG: last completed transaction was at log time
>> 2010-06-03 08:11:17.531-04
>> 2011-06-03 08:46:56 EDTFATAL: xlog flush request 18/A2BE3510 is not
>> satisfied --- flushed only to 9/8E080968
>> 2011-06-03 08:46:56 EDTCONTEXT: writing block 0 of relation global/1261_vm

>> How do we recover from this?

> Reset the time to be correct.

I don't think that failure has anything to do with system clock time.
It looks more like filesystem corruption or missing pg_xlog files.

            regards, tom lane

My thanks for your replies.  We used pg_resetxlog to clear things up.
The database was not in active use, so the loss of the transactions
didn't matter.

RobR