When I said earlier about multipath bugs, I really meant RH4 and anything older than RH 5.8, especially if you have active -passive configs with round robin and SAN dual controllers. We actually have more problems with NetApp than 3par in this area. Are you using thin-provisioning with 3par ? That's a nice feature and really works, but be careful about tuning the 3par arrays and don't overcommit your I/O rates or number of LUN's. There may still be a limit on the 8192 LUN count, which includes Snaps as well. That said, 3par works pretty darn good- a lot of people didn't believe in the whole thin-provisioning concept would even work (i.e. smoke and mirrors) when we started with it 6 years ago. -----Original Message----- From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Martin Bach Sent: Friday, July 27, 2012 4:20 AM To: oracle-l@xxxxxxxxxxxxx Subject: Re: Linux Native Multipath, ASM and Instance Failures Jon, On 26/07/2012 23:08, CRISLER, JON A wrote: > you want to check your multipath.conf to make sure it has all the > right options that's probably the most important point: check your vendor documentation to ensure that the devices {} are set according to what you need. There are subtle differences in the way that I/Os are distributed, what the library does when it can't write to a device, and which method should be used for path checking. The differences exist even with different product lines of the same vendor: VNX and Symmetrix for example have different entries in the devices {} section! I'm sure the same applies to any other storage product. There are metalink notes on how to set up multipath.conf in RHEL 5 but although broadly speaking they are correct, they are not necessarily optimal for your configuration. What's of importance when troubleshooting this kind of problem are kernel (uname -a), and the dm-multipath RPM version. There were known issues in 5.5 but they have alledgedly been fixed in 5.6 and later. Hope this helps, Martin -- //www.freelists.org/webpage/oracle-l -- //www.freelists.org/webpage/oracle-l