Re: Hung postmaster (8.3.9)

["Ed L." <pgsql(at)bluepolka(dot)net>]

On Saturday 27 February 2010 @ 22:18, Greg Smith wrote:
> Ed L. wrote:
> > 2010-02-27 20:11:10.426 CST [23134]    LOG:  database system
> > was not properly shut down; automatic recovery in progress
> > 2010-02-27 20:11:10.497 CST [23134]    LOG:  record with
> > zero length at 8BC/63F6020 2010-02-27 20:11:10.497 CST
> > [23134]    LOG:  redo is not required
> 
> If the server is hung at this point, it was doing something: 
>  recovery from a crash, and during that period it will not
>  respond to statements until it's finished as you noted.  That
>  can take some time, could run into the minutes.  And if you
>  kill it, next time the process starts all over again having
>  made no progress.
> 
> I'd try starting the server again, confirm it's only after
>  this point it pauses at, watching if it's doing something
>  with "top -c", and grab some backtraces of what it's doing
>  then.  In this sample you gave, the shutdown was too fast for
>  the server have done very recovery before it was killed off. 
>  It may just be you need to let it sit there for a while to
>  finish recovery and then it will be fine again, can't tell
>  from the data you've provided so far.
> 

Good point from my example, but it hung like that for 8 hours, 
and seems to be ready to do so again.