Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Wed, 2008-09-03 at 09:45 +0200, Lars Ellenberg wrote: > btw, when you have a non-volatily, > battery backed up write cache in the controller, > don't forget to disable the (typically volatile!) on-disk cache. > or you may face data loss on power outage anyways. > you may enable the read cache on disk, if you think it is worth it. > > in this setup, that was done by > modprobe sg > for i in `seq 4`; do sdparm -s WCE=0,RCD=0 -S /dev/sg$i ; done > > see man sdparm for details, > and your controller manual for other ways to achieve the same > if the single disks are not visible as sg devices to the OS. I have done some tests with a very nice perl script from http://brad.livejournal.com/2116715.html with different RAID controllers (Adpatec, Areca, 3ware, and SATA-onboard) and documented the results (in German) here: http://www.thomas-krenn.com/de/wiki/Cache_Einstellungen_von_RAID_Controllern_und_Festplatten#Einstellungen_f.C3.BCr_sicheren_Betrieb One of the controllers I've tested was an Adaptec 5405 - with this controller you can use the arcconf CLI tool from Adaptec to disable the cache of a disk, e.g. for the first disk: arcconf SETCACHE 1 DEVICE 0 0 wt PS: I've also done some consistency tests and provided some infos on that, too: http://www.thomas-krenn.com/de/wiki/Konsistenz_des_RAIDs_%C3%BCberpr%C3%BCfen best regards, Werner