Struggling with I10CHR raid

[KentState]

For over a year, I successfully ran a RAID 5 setup using Intel RE3 1TB disks. Recently, I picked up another drive and rebuilt it as RAID 10. Since then, two of the older disks and the new one keep reporting as a failure. I'm able to rebuild and everything is ok for a day, then the same drives start randomly failing.

Should I:

1. RMA the defective drives and assume that they are the problem.
2. Is RAID 10 the problem and should I go back to RAID 5?
3. Buy some other PCI-E like http://www.newegg.com/Product/Produc...82E16816115050.

The system runs Win7-64, ASUS P6T Deluxe V2, i7-920, 12 GB RAM, Vortex SSD, 4 RE3 WD 1TB. I run VMWare Workstation and use the RAID set for the Virtual Machines. I would like to keep RAID if possible. Let me know if you need any other information.

 

[RebateMonger]

I don't know why you are having problems, but I'd certainly be cautious. With that kind of failure rate, you could easily have a second dropped disk while in Rebuild mode and quite possibly lose the entire array. Do you keep backups?

That's pretty strange to be having problems with three out of four disks. Usually it'll be a single disk and either the disk isn't responding quite right or there's a cabling problem. But three disks?

Sorry I'm not providing much help. Actually, I am interested in your performance readings from that RAID 10 array on ICH10R. Do you have any Read/Write numbers? Do you have any numbers from those same disks on RAID 5? I know you have other things to worry about right now, but I am curious.

 

[KentState]

I have the contents of the array backed up to an external drive as well as about 40GB through Mozy. The VM's themselves are not backed up inside so any changes would be lost.

Can you point me to a free disk benchmarking tool? I could take a reading in the current RAID 10 configuration and take another with RAID 5 if the opportunity presents itself. I know that sequential read/write performance is very high when transferring to/from the SSD. It is also fast enough to house 6-8 VM's running with 2 of those having SQL Server 2008 with a 100GB database. The disk queue length stays below a manageable 10 and closer to 2-3 at most times.

 

[RebateMonger]

I really don't spend much time benchmarking. There are likely better programs, but I usually just use HDTune. I use the free version, but that version only benchmarks "Reads". But you could download the "Pro" version, that also measures "Writes" and use it for a while.

Yeah, I WOULD like to know what your readings are. In the meantime, hopefully I or somebody else can give you some more help with your array stability.

 

[RebateMonger]

Are there any entries in the Windows System Event Logs from the Intel Storage Management program (whatever it's called nowadays)?

 

[FishAk]

Since then, two of the older disks and the new one keep reporting as a failure.

Did you happen to install the new drive between the two older ones that give you problems now? If so, it may be that the close proximity of the drives allows inadequate air circulation, and now the drives are getting too hot- causing problems.

 

[KentState]

Did you happen to install the new drive between the two older ones that give you problems now? If so, it may be that the close proximity of the drives allows inadequate air circulation, and now the drives are getting too hot- causing problems.

The new drive is on the second row and has nothing next to it.

 

[taltamir]

Intel RE3 1TB disks

do you mean WD RE3 disks?

Also, when you say "since then" you mean the same/next day you started having problems? a week later? a month later?

 

[Emulex]

things you don't do with raid5:
1. crash - not allowed.
2. lose power - not allowed (intel ich10R)
3. reset button - just straight up retarded.
4. emergency shutdown - make sure your drives settle. i actually would prefer a timeout like 10-20seconds before a shutdown occurs - esp if you enable the drive cache. Think PRE-ACPI "you may now power off your pc" - far better than having a dirty shutdown.


with or without a real raid controller those rules pretty much apply.

 

[KentState]

do you mean WD RE3 disks?

Also, when you say "since then" you mean the same/next day you started having problems? a week later? a month later?

I was thinking about the controller when I said Intel, but meant WD RE3. It was within a few days after adding the new drive and changing it to RAID 10.

 

[KentState]

BTW, RebateMonger, here is the comparison of the RAID10@64k vs my OZC Vertex 120GB SSD. I ran this for a 1GB sequential file. I don't have a tool that can measure random I/O and HDTune Pro is not free unless I missed something.

SSD

RAID10

I'm pretty happy with the performance.

 

[Emulex]

AS benchmark and CDM 3.0 are commonly used since they send random data. ATTO is not a good choice for sandforce controllers since they have compression to skew benchmarks (wonder why they dont use this tech on hard drives?).

problem is how are you going to get 6TB of SSD for $500? $5000? 50000?

 

[taltamir]

I was thinking about the controller when I said Intel, but meant WD RE3. It was within a few days after adding the new drive and changing it to RAID 10.

the simplest way to know for sure is to go back to raid5 and see if it still does that.