[ Date Index ][
Thread Index ]
[ <= Previous by date /
thread ]
[ Next by date /
thread => ]
Bill Wilson wrote:
RAID 5 OK but serios degradation will take place during a failure.
Modern drives are so reliable failure tends to be due to external issues (mostly operators bumping into disk arrays and the like) with these failure models you assume more multiple disk failures, although I have to say I've haven't seen any multiple disk failures in the same RAID group yet<!>.
Raid 1 will not degrade the same way in the event of failure. However at least 50% of our failures are of the RAID card itself or the controller in software RAID.
Eek - what RAID cards are these to avoid ? ;-) We saw a couple of failures with clients failing to replace broken disks in RAID-5 arrays at my previous place, so when the second disk went.....
If you are concerned about this the only solution is RAID 10 with each pack on a separate controller or at least a different channel on the same controller.
Anyone done this on Linux? HP-UX does this really well, you just cable in an appropriately configured disk array, fire up disk management software, and it say (something like) "I have found a second path to an existing volume, should I use this in the event of an I/O channel failure Y/N?" (I wonder how many people ever select "N" at this point ?! - Actually I know for sure someone at Cap Gemini managed not to switch this feature on for a client of mine - a case of read the manual and THEN speak to an HP field engineer - they even had the same configuration mistakes as were in the HP manual!). Of course if you want to load balance across the I/O channels (and you usually do), you still have to bail out to the command line afterwards, and delete the primary channel, and do the same procedure again <I guess that is an 9 out of 10 for usuability ;-> I've nearly always used the Data General Clariion disk arrays for this in the past, but they are kind of past their sell-by-date. They support redundant controllers, with an auto-trespass feature, so will do the full monty of failing between channels etc. Although it takes one machine I look after 2 minutes to completely fail over all it's Oracle volumes when you pull the SCSI cable out ;-). One I haven't used is the HP arrays with hot-spot(?) technology, as I understand it, this is a kind of adaptive RAID, where it try to tune the style of redundancy to match the use the machine is making of it's disks. Anyone any experience of these? Simon, trying to work around an Oracle "export" bug, which is failing after about 4 hours zzzZZzzzZZZ, geez there were a lot of export fixed in 8.1.7, that weren't in 8.1.6 :-(, I really don't want to be upgrading Oracle just(!) to defragment a database. -- The Mailing List for the Devon & Cornwall LUG Mail majordomo@xxxxxxxxxxxx with "unsubscribe list" in the message body to unsubscribe.