During a shutdown immediate, the LGWR and DBWR must write there log buffer
and buffer pool to the datafiles. If your buffer pool is large and the
checkpoint interval is quite large, there may be many 'dirty' blocks in the
buffer pool. They all has to be written to the datafiles. This may take some
while...
Are there active sessions at the moment of shutdown? Not all kind of
queries/data manipulations can be 'killed' immediate. This may also cause
the 'problem'.
"Anton Dischner" <nospam_at_nospam.org> wrote in message
news:030620031512220253%nospam_at_nospam.org...
> Hi,
>
> every week we do a nightly switch over between our master/slave
> configuration via an automated script.
>
> Sometimes the shutdown immediate hangs for no reason without any
> log-entries.
>
> I believe this is a bug but. Patching from 8.1.7 to 8.1.7.4 didn't
> change the behaviour.
>
> shutdown abort is a solution but not a good one.
>
> Any more ideas?
>
> SuSE 7.1, kernel 2.2.18-SMP, 4 * Xeon, 2GB main memory
> all programs stopped at this time.
>
> PS: I followed the shutdown from root user discussion ;-)
Received on Tue Jun 03 2003 - 15:47:06 CDT
