betcha its the crap controllers you are always buying :-)

Since my beginning of doing computers for work... purchasing up to $10,000 (6 channel caching) disk controllers and Multi-path setups costing even more because of the SAN back end.

Sadly Box... we are talk 4 out of every 5 disk failure events result in losing the array... be it a Mirror, Duplex, a RAID5 or RAID50 or a SAN...

I've lost mirrors due to corruption caused by the controller writing bad data due to the "failed disk" causing the controller to screw up.

On Duplex setups... lets just say Garbage In, Garbage Out.

I've lost Raid 5 Arrays multiple times due to multiple simultaneous disk failures (even on distinctly different ages/manufacturers of drives).

On RAID50, I lost a controller and the newest driver that was installed on the system before the loss (on AIX), caused the whole shelf to be lost.

The Multi-path SAN stuff... we lost multiple storage blocks due to a single drive error and even lost the FLASH SNAPSHOTs of the storage blocks.

I am just not at all impressed with Hardware RAID controllers or even these expensive SAN systems.

Just so you know, the reason the ratio is 4 out of 5, we had a NetAPP FAS250 that had multiple failures of disks... never once was it out of service due to the disk failures. We swapped the drives and assigned them to the unit... voila. Worked good as ever. The only problem we really had with the FAS250 was non-redundant heads. We had a hiccup a few years back once. Replacement took more than 6 hours... (guaranteed to be less than 4 hours). After that event we found out that we could have had redundant heads... but due to the age of the machine NetAPP would not sell us an additional one and would not support us buying a used one... feh.

Buhbye NetAPP.

As far as *I AM* concerned Linux's MD RAID setup has only ever bitten me once. That was with my personal 600MHz file server. I had two arrays of 4 disks each appended together. I had two MD (md0 and md1) they had PVs one them... I added them all to a single LV. I lost a single drive in the older set of disks, which meant the hot spare should have kicked in... This being a 2.4.22 kernel and I had not updated in forever to get a new one... it never "noticed" the dropped drive to begin the rebuild... It just dropped the data and never rebuilt it... and before I noticed it was to late.

I lost nearly 33% of my music (read errors resulting in zero length files) and almost 100% of video and XBOX game images (same zero length thing)

Of course, I had no backups for my personal shi^H^Htuff. 3.4TB back a few years ago, was insane to have backed up for personal stuff.

Post #337,782 by boxley 12/28/10 9:21:58 AM Reply	betcha its the crap controllers you are always buying :-) Any opinions expressed by me are mine alone, posted from my home computer, on my own time as a free American and do not reflect the opinions of any person or company that I have had professional relations with in the past 55 years. meep
Post #337,783 by folkert 12/28/10 9:49:41 AM Reply	Sadly... NO. Since my beginning of doing computers for work... purchasing up to $10,000 (6 channel caching) disk controllers and Multi-path setups costing even more because of the SAN back end. Sadly Box... we are talk 4 out of every 5 disk failure events result in losing the array... be it a Mirror, Duplex, a RAID5 or RAID50 or a SAN... I've lost mirrors due to corruption caused by the controller writing bad data due to the "failed disk" causing the controller to screw up. On Duplex setups... lets just say Garbage In, Garbage Out. I've lost Raid 5 Arrays multiple times due to multiple simultaneous disk failures (even on distinctly different ages/manufacturers of drives). On RAID50, I lost a controller and the newest driver that was installed on the system before the loss (on AIX), caused the whole shelf to be lost. The Multi-path SAN stuff... we lost multiple storage blocks due to a single drive error and even lost the FLASH SNAPSHOTs of the storage blocks. I am just not at all impressed with Hardware RAID controllers or even these expensive SAN systems. Just so you know, the reason the ratio is 4 out of 5, we had a NetAPP FAS250 that had multiple failures of disks... never once was it out of service due to the disk failures. We swapped the drives and assigned them to the unit... voila. Worked good as ever. The only problem we really had with the FAS250 was non-redundant heads. We had a hiccup a few years back once. Replacement took more than 6 hours... (guaranteed to be less than 4 hours). After that event we found out that we could have had redundant heads... but due to the age of the machine NetAPP would not sell us an additional one and would not support us buying a used one... feh. Buhbye NetAPP. As far as I AM concerned Linux's MD RAID setup has only ever bitten me once. That was with my personal 600MHz file server. I had two arrays of 4 disks each appended together. I had two MD (md0 and md1) they had PVs one them... I added them all to a single LV. I lost a single drive in the older set of disks, which meant the hot spare should have kicked in... This being a 2.4.22 kernel and I had not updated in forever to get a new one... it never "noticed" the dropped drive to begin the rebuild... It just dropped the data and never rebuilt it... and before I noticed it was to late. I lost nearly 33% of my music (read errors resulting in zero length files) and almost 100% of video and XBOX game images (same zero length thing) Of course, I had no backups for my personal shi^H^Htuff. 3.4TB back a few years ago, was insane to have backed up for personal stuff.
Post #337,787 by malraux 12/28/10 10:00:47 AM Reply	You get what you pay for with SAN. The only stuff worth anything is the high-end: EMC, some Hitachi. Regards, -scott Welcome to Rivendell, Mr. Anderson.
Post #337,790 by jake123 12/28/10 11:25:57 AM Reply	And Sun we've had very good results in our systems with drive failures failing to take down the raid. The one time we did have a real problem it turned out that a fire in the ventilator unit right above our rack had resulted in coating the inside of the raid with soot. We brought it back, cleaned it up, and put it in the lab where it's still working fine. We /could/ put it back into production, but as a matter of policy once there's one failure the hardware just doesn't go back into production, no matter what.
Post #337,824 by malraux 12/29/10 12:36:48 PM Reply	Sun is rebadged Hitachi gear Regards, -scott Welcome to Rivendell, Mr. Anderson.
Post #337,796 by boxley 12/28/10 6:25:18 PM Reply	actually Compellant, 3PAR and Pillar are equal or better than EMC and usually cost less Any opinions expressed by me are mine alone, posted from my home computer, on my own time as a free American and do not reflect the opinions of any person or company that I have had professional relations with in the past 55 years. meep
Post #337,825 by malraux 12/29/10 12:37:20 PM Reply	Financial services All things being equal, apart from cost, go with the name. Regards, -scott Welcome to Rivendell, Mr. Anderson.

Welcome to IWETHEY!