Rocketraid 2300 Raid 5 4x2TB causing 100% system utilization, then crash

[CptCrunch]

Some backstory. Had this card RocketRaid 2300 on a Dell Inspiron 530 Win 2k8 r2 x64 server with a Raid5 4x500gb Seagate drives and it had 100% uptime. Time comes to update the server to: Q6600, EVGA 570SLI motherboard, 8gb DDR2, 74gb raptor, etc (old system) and replace the 500gb drives with a Raid5 4 x 2TB Samsung drives. Upgrade happened all at once. I did add a 600 watt coolermaster powersupply at this time, previously unused.

So the raid array is not stable at all. Sometimes it'll handle an un-rar across network drives, or extraction of a large (several GB) nzb download, and other times it'll die when transferring a few hundred megabytes across the network. When the system "dies" it does this; in process explorer the SYSTEM goes to 25% utilization, and then the system is very slow. No RDP, no access to network shares, DHCP fails to assign IP addresses, etc. The only solution is to restart the machine.

I've tried changing the raid card's write back policy, and tried to reduce the load on the raid card, but nothing seems to work. I'm not sure if its the motherboard, (I've had several of these EVGA boards replaced due to various issues,) or if the raid card can't handle a 8TB Raid 5 array. At this point I have no problem buying a new raid card or motherboard, whichever will fix the issue.

 

[Emulex]

buy a real hardware raid solution. i have a few LSI raid controllers (BR10i) with 8 port fanout cables for $50. those software raid controllers suck.

next: use raid edition drives.
next: use raid-10 (unless you have a BBWC/FBWC you will be in large pain with raid-5)

I tend to use servers (ml110) with ecc and xeon processors since they are designed for end to end ECC. They are cheap and if they are setup to be compatible with vmware esxi HCL they are a solid tested design.

could be a bad set of drivers or too much junk sitting on one IRQ that causes the sporadic failures. I've given up on consumer raid solutions after many failures over the long run. Just not worth the frustration and anxiety.

what does eventvwr show? On my servers i see 0 warnings and criticals over a period of years.

Windows 2008 R2 has built in memory test - might give that a shot too - how are you mitigating the heat? you have the cpu profile to 100% all the time - no sleep/hibernation? the inspiron 530 had a terrible nic on board. not sure how good the EVGA solution is.

Also if your powersupply is high efficiency you may get unstable when a cheap square wave batter backup kicks in - puresine ftw. Test is simple. put the battery on backup mode and run some memory tests and or prime95 and look for corruption. had this happen to me recently with a new efficient power supply. as soon as it kicked to the cheap be550 it would corrupt and crash intermittently on battery power. replaced it with a cyberpower pure-sine and the problems went away. very repeatable.

 

[CptCrunch]

Thanks for the post Emulex. I did some testing and I apparently have bad ram. The ram wasn't bad when I tested it with Windows 2008's built in tester, but Memtest86 said it was bad within the first 3 seconds. So I pulled those 2 sticks and the system has been rock solid stable since. I would like to get a better hardware raid card, so I think i'll be looking for one soon.