Unsolved
This post is more than 5 years old
2 Posts
0
4205
data corruption on Perc 5/i with RAID 5
I have two PE 840 servers (running CentOS 5), one having two SATA disks defined on the controller as mirroring, and the second with 4 SATA disks configured as RAID 5. I have experienced a few cases of bad data corruption on this second (RAID) server when there's a power outage: when the UPS runs out of juice and the servers go off - i.e. not a clean shutdown. Both servers have their write cache setting as "write back" (this controlller has a battery). The first server (mirroring) never has had any problems recovering from these crashes, but the second (RAID) does, repeatedly.
Beyond the obvious solution of a better UPS with automated shutdown, I'd like to ask:
First: is this expected behavior?
Second: Can I change the write cache policy to "Write Thru" on a running server, without loosing the RAID configuration and without loss of data? Will this help mitigate the problem any?
Thanks,
Micha
Dev Mgr
4 Operator
4 Operator
•
9.3K Posts
0
November 5th, 2009 06:00
This shouldn't happen, as that's what the battery is supposed to be for, but power outages always pose a risk.
Do you have an option to have the servers shut down gracefully if the power is out for more than a few minutes (ie. does your UPS have a way to alert your server if the power is out for more than a certain period of time)?
As for changing the cache policy, assuming you have OpenManage installed, you can change that setting on the fly (not sure if there's a command line option if you're wanting to change this only when "the UPS kicks in").
michasilver
2 Posts
0
November 5th, 2009 07:00
Thanks for replying.
Yes we will certainly get the UPS setup to call a proper shutdown after x minutes. You mention that I can change the cache policy on the fly. Is the below screen shot the Open Manage option I'm looking for?
Thanks again,
Micha