Unsolved
This post is more than 5 years old
25 Posts
0
39478
October 21st, 2005 17:00
Linux reports cache data is unavailable and must then assume write through
Hi, I've installed Suse SLES 9 XT64 on a PowerEdge 1850 with a Perc4/I RAID and a QLogic QL2340 connected to an EMC AX100 (single controller) through a Fibre channel switch.
When the server boots, I get the following 'error' for both the Perc4/I and the AQL2340
sda: asking for cache data failed
sda: assuming drive cache: write through
sdb: cache data unavailable
sdb: assuming drive cache: write through
Server is operational but I would get better performance if I could activate the cache write back. Both are configured in 'write back'. What could be wrong for this to happen?
Thanks.
When the server boots, I get the following 'error' for both the Perc4/I and the AQL2340
sda: asking for cache data failed
sda: assuming drive cache: write through
sdb: cache data unavailable
sdb: assuming drive cache: write through
Server is operational but I would get better performance if I could activate the cache write back. Both are configured in 'write back'. What could be wrong for this to happen?
Thanks.
No Events found!
ctacat
6 Posts
0
October 29th, 2005 12:00
This message comes from scsi disk (or adapter ?) driver, which may try to guess what type of write cache is used on the disk : write-through or write-back. It's probably able to adapt its behaviour based on this guess (I don't know how, I don't have time to read scsi disk driver source code ;-) Maybe some real expert could answer here).
The SCSI controller shipped with Poweredge (frequently RAID controller) is probably not able to send a good reply to scsi disk driver to permit it to guess correctly. So it fallbacks to write-through.
However, I already keep a small doubt : is it risky to let scsi disk driver believes the cache is write-through, when the scsi controller configured its volumes as write-back ?