Thanks to (Alan Rollow - Dr. File System's Home for Wayward Inodes.) who
replied to my comment about relying on UPS for enabling Write back cache on
RAID controlloer KZESC.
Alan wrote :
>re: Trusting a UPS.
>
>Do you have a rigorous plan in place to periodically test the UPS
>to make sure it really is uninterruptable? If not, then I wouldn't
>trust the write back cache on a RAID to a UPS. UNIX systems have
>been using a host based write back cache almost since the beginning.
>Early UNIX file systems were very fragile because file system meta-
>data was treated the same as data and cache. A power failure could
>severly damage a file system. Today's implementation typically write
>file system metadata synchronously to ensure consistency, even if
>ordinary data may lost.
>
>But an XOR based RAID is special. A particular block of parity affects
>all the other data in particular strip (row, slice, etc). If you lose
>the parity, then not only will the data being written with be affect
>but data that may have been stable for ages. If you also lose a
>device in the power failure (not uncommon) and have to regenerate
>the data all sorts of things could be broken.
>
>That is why you have to hold a UPS particularly suspect when it
>providing power to a RAID with writeback cache.
>
Received on Sun Nov 05 1995 - 21:11:08 NZDT