Hi,
currently we are running an oracle db on an FAS6070c and today we had some complaints that for some hours operations which use a lot of getattr-ops, like ls -la does, are slower than they used to be.
An oracle-colleague told me that he thinks there is a best-pratcise for volumes with oracle-db inside which describes something that boosts up getattr's.
I did not find anything appropriate on NOW. Has anybody ever come accross this issue?
Best Regards and thanks in advance
Jochen
On 12/5/06, Willeke, Jochen Jochen.Willeke@wincor-nixdorf.com wrote:
An oracle-colleague told me that he thinks there is a best-pratcise for volumes with oracle-db inside which describes something that boosts up getattr's.
vol options <vol name> no_atime_update on
do i have to check some things before activating this? I mean is it absolutely for sure that the DB does record it's own timestamps?
Rgds
Jochen
-----Original Message----- From: Sto Rage(c) [mailto:netbacker@gmail.com] Sent: Tuesday, December 05, 2006 7:13 PM To: Willeke, Jochen Cc: toasters@mathworks.com Subject: Re: optimize volume/qtree for oracle db
On 12/5/06, Willeke, Jochen Jochen.Willeke@wincor-nixdorf.com wrote:
An oracle-colleague told me that he thinks there is a best-pratcise
for
volumes with oracle-db inside which describes something that boosts up getattr's.
vol options <vol name> no_atime_update on
This is only the read access time, not the write/create time. <scw> On Tue, Dec 05, 2006 at 08:11:21PM +0100, Willeke, Jochen wrote:
do i have to check some things before activating this? I mean is it absolutely for sure that the DB does record it's own timestamps?
Rgds
Jochen
-----Original Message----- From: Sto Rage(c) [mailto:netbacker@gmail.com] Sent: Tuesday, December 05, 2006 7:13 PM To: Willeke, Jochen Cc: toasters@mathworks.com Subject: Re: optimize volume/qtree for oracle db
On 12/5/06, Willeke, Jochen Jochen.Willeke@wincor-nixdorf.com wrote:
An oracle-colleague told me that he thinks there is a best-pratcise
for
volumes with oracle-db inside which describes something that boosts up getattr's.
vol options <vol name> no_atime_update on
Sto Rage(c) [mailto:netbacker@gmail.com] wrote: [...]
vol options <vol name> no_atime_update on
On Tue, Dec 05, 2006 at 08:11:21PM +0100, Willeke, Jochen wrote:
do i have to check some things before activating this? I mean is it absolutely for sure that the DB does record it's own timestamps?
Stephen C. Woods writes
This is only the read access time, not the write/create time.
It's much more likely that it's the other clients sharing the volume, not the Oracle DB, which will be (mildly) perturbed by the absence of atime updates ("mildly" because client-side caching means atimes have dodgy semantics under NFS in the first place).
If you can put your Oracle database files in a separate (maybe flexible) volume, then you can turn off atime updates just for that volume.
Jochen I don't have the answer to your question but I do have a question for you. We have an Oracle db here that the dba wants to run it on an EMC system using RAID 10. We would have to buy the system but since we already have NetApp I am trying to convince him that a NetApp system (FAS960 or FAS3050) would do the job just fine. Can you let me know how you have your Oracle on NetApp configured and how well it works for you?
Thanks Mike Miller
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Willeke, Jochen Sent: Tuesday, December 05, 2006 11:01 AM To: toasters@mathworks.com Subject: optimize volume/qtree for oracle db
Hi,
currently we are running an oracle db on an FAS6070c and today we had some complaints that for some hours operations which use a lot of getattr-ops, like ls -la does, are slower than they used to be.
An oracle-colleague told me that he thinks there is a best-pratcise for volumes with oracle-db inside which describes something that boosts up getattr's.
I did not find anything appropriate on NOW. Has anybody ever come accross this issue?
Best Regards and thanks in advance
Jochen
Hi,
we had ORACLE DB's running on FAS3050 for a long time. The only bottleneck we got were high RTT's in some cases but we managed to better this with bigger aggregates. I think there is a recomendation not to fill WAFL more than 80% as it is true for every filesystem i know.
The throughput was always ok on our DB-volumes. If your dba is not totaly convinced you can offer him the use of 15k-drives. We use them too and got great RTT's (normaly about 2-3ms) on our db-volumes.
I have never seen an FAS960 running oracle, but once i talked to a consultant from oracle and he said, that in his lab he has some 960 running oracle db's over NFS and it works fine.
On argument you can offer is that oracle itself uses Netapp and NFS and runs one of the biggest Netapp-Instalations in the world in their datacenter in Austin.
Another good point might be the capabilities of FlexClone. We do not use it right now but it can speed up development and integration! This is not special for 3050 or 960 but it is a plus for netapp.
For configuration issues you can find some best-practises and other articles on NOW. I would suggest to read them first, because there are some important points about mount-options! But i am still no expert about this whole topic, but i do my best to become one ;-)
Regards
Jochen
-----Original Message----- From: Miller, Michael CTR USTRANSCOM J2 [mailto:michael.miller.ctr@ustranscom.mil] Sent: Tuesday, December 05, 2006 7:38 PM To: Willeke, Jochen; toasters@mathworks.com Subject: RE: optimize volume/qtree for oracle db
Jochen I don't have the answer to your question but I do have a question for you. We have an Oracle db here that the dba wants to run it on an EMC system using RAID 10. We would have to buy the system but since we already have NetApp I am trying to convince him that a NetApp system (FAS960 or FAS3050) would do the job just fine. Can you let me know how you have your Oracle on NetApp configured and how well it works for you?
Thanks Mike Miller
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Willeke, Jochen Sent: Tuesday, December 05, 2006 11:01 AM To: toasters@mathworks.com Subject: optimize volume/qtree for oracle db
Hi,
currently we are running an oracle db on an FAS6070c and today we had some complaints that for some hours operations which use a lot of getattr-ops, like ls -la does, are slower than they used to be.
An oracle-colleague told me that he thinks there is a best-pratcise for volumes with oracle-db inside which describes something that boosts up getattr's.
I did not find anything appropriate on NOW. Has anybody ever come accross this issue?
Best Regards and thanks in advance
Jochen