Can anyone tell me what this message means. I am running Opensuse 10.3.

Jul 23 09:29:17 hplinuxserver kernel: sd 8:0:0:0: [sda] Assuming drive cache: write through
Jul 23 09:29:17 hplinuxserver kernel: sd 8:0:0:0: [sda] Assuming drive cache: write through
Jul 23 11:05:13 hplinuxserver kernel: sd 9:0:0:0: [sda] Assuming drive cache: write through
Jul 23 11:05:13 hplinuxserver kernel: sd 9:0:0:0: [sda] Assuming drive cache: write through

It's from my Warn Logfile and I would like to fix it before the server goes live.

Better late than never (i.e. almost two years later), but since this came up second in a Google search...

...I worried myself, especially since I had a bout with my machine: freezing on reboot (getting error "could not stat the resume device file"), but not on a hard boot. Uninstalling uswsusp helped my case, but I still have the "Assuming drive cache: write through" (and "Write Protect is off"), but I don't seem to have any issues. My machine is up and running for three years, if that helps.

Considered "dd if=/dev/urandom of=/dev/sdx" or "fdisk /dev/sdx" (and redo (all) the partition(s) and "reboot" and "mkfs.ext3 /dev/sdx"), but since working, didn't rock the boat.

I do have a Dell and found this link on another forum:
http://lists.us.dell.com/fom-serve/cache/110.html

But this also happens for a firewire (or USB) attached external 8TB drive (that is not attached to that RAID controller), yet that drive registers a "Attached SCSI disk"; so, linux sees all disks as SCSI and can't handle some identifier on any disk to help it "know" and not "assume"?! Guess I could get some experience and install Linux on a machine that would definitely have /dev/hdx (and not /dev/sdx).

Bottom line: I have no idea, but have been living with the message (which could just be information) and all is fine (but I'm with you on the nagging feeling that all should be in harmony for a production machine).

This thread won't die!

I'm currently cleaning a hard disk drive and besides I'm checking an USB pendrive. I know this message and I was never worried about this. But what I don't understand - when the kernel assumes a cache why does it use the write through method? The assumption and the behaviour doesn't make sense in my optinion.

• don't assuming write cache : write through <- sounds sensible
• assuming write cache : anyway, write through <- that's still all right
• assuming write cache : write back <- perfect
• assuming write cache : write trough <- isn't a logical conclusion