Hello Tony, If the hardware hasn't changed I'd look at the workload on the database server. If the customer is taking regular statspack snapshots they might be able to see whats causing the extra activity. They can use AWR or the diagnostic pack, if they are licensed, to see the offending SQL or PL/SQL or any hot objects.
However if you want to tune at the ZFS level then the following has some advice for ZFS and databases http://www.solarisinternals.com/wiki/index.php/ZFS_for_Databases. On 9 December 2010 15:48, Tony Marshall <tony.marsh...@oracle.com> wrote: > Hi All, > > Is there a way to tune the zfs prefetch on a per pool basis? I have a > customer that is seeing slow performance on a pool the contains multiple > tablespaces from an Oracle database, looking at the LUNs associated to that > pool they are constantly at 80% - 100% busy. Looking at the output from > arcstat for the miss % on data, prefetch and metadata we are getting around > 5 - 10 % on data, 50 - 70 % on prefetch and 0% on metadata. I am thinking > that the majority of the prefetch misses are due to the tablespace data > files. > > The configuration of the system is as follows > > Sun Fire X4600 M2 8 x 2.3 GHz Quad Core Processor, 256GB Memory > Solaris 10 Update 7 > ZFS Arc cache max set to 85GB > 4 Zpools configured from a 6540 Storage array > > - apps - single LUN (raid 5) recordsize set to 128k, from the array, > pool contains binaries and application files > - backup - 8 LUNs (varying sizes all from a 6180 array with SATA disks) > used for storing oracle dumps > - data - 5 LUNs (Raid 10 6 physical drives) recordsize set to 8k, used > for Oracle data files > - logs - single LUN (raid 10 from 6 physical drives) recordsize set to > 128k, used for Oracle redo log files, temp db, undo db and control files. > > 18 Solaris 10 zones, of which 12 of these are oracle zones sharing the data > and logs pools. > > I think that the prefetch will be useful on the apps and backup pools, > however I think that on the data and logs pools this could be causing issues > with the amount of IO that is being caused by the prefetch and the amount > that it is missing in the arcstats could be the reason why the devices are > at 100% busy. Is there a way to turn the prefetch off for just a single > pool? Also is this something that can be done online or will it require a > reboot to put into effect. > > Thanks in advance for your assistance in this matter. > > Regards > Tony > -- > [image: Oracle] <http://www.oracle.com> > Tony Marshall | Technical Architect > Phone: +44 118 924 9516 <tel:+44%20118%20924%209516> | | | Mobile: +44 > 7765 898570 <tel:+44%207765%20898570> > Oracle Remote Operations Management > United Kingdom > > ORACLE Corporation UK Ltd is a company incorporated in England & Wales | > Company Reg. No. 1782505 | Reg. office: Oracle Parkway, Thames Valley Park, > Reading RG6 1RA > [image: Green Oracle] <http://www.oracle.com/commitment> Oracle is > committed to developing practices and products that help protect the > environment > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss > > -- Thanks A Jabbar Azam
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
