-1

So this is the second time in 2-3 months that my my dedicated box out at a unnamed host has had faulty memory in some shape or form. Is this normal or is my host just giving me cheap memory sticks? How can I find out?

Memory Issue 1:

EDAC MC0: UE row 0, channel-a= 2 channel-b= 3 labels "-": (Branch=1 DRAM-Bank=7 RDWR=Write RAS=32679 CAS=0 FATAL Err=0x4 (>Tmid Thermal event with intelligent throttling disabled))

Memory Issue 2:

 EDAC MC0: UE row 0, channel-a= 0 channel-b= 1 labels "-": (Branch=0 DRAM-Bank=0 RDWR=Read RAS=0 CAS=0, UE Err=0x2 (Aliased Uncorrectable Non-Mirrored Demand Data ECC)) 

Specs:

  • Dual Intel quad-core server with Intel 5400 Series Harpertown processors
  • Up to 24GB DDR2-FBDIMM 667Mhz memory
  • CentOS 5.6 (Final)
5
  • 1
    I'm assuming you've figured out that your unnamed memory error is related to hardware. It sounds more like a customer support issue for your unnamed hosting provider.
    – Nixphoe
    Sep 2 '11 at 13:53
  • 4
    If you're really that bothered about the downvote, it was mine. There is literally nothing we can do to help you. You need to take this up with your host, and if you're not satisfied then escalate it with them and make a big fuss. If you're still not satisfied, demand a refund and take your business elsewhere. We can help with absolutely none of those things. Sep 2 '11 at 13:56
  • 5
    You should ask them if it's cheap memory, and you need to trust what they tell you. If you don't trust their word, then move. Again, I'm sorry but we can't help you. Sep 2 '11 at 14:00
  • Have to agree with @BenPilbrow. There's no way to tell really what you're running on a managed provider without just asking them. I suppose you can query your hardware and try to ascertain if the hardware is "quality" or not in general, but you're still shooting in the dark. Your best bet is to call them and tell them that you've had X failures in Y months and you're getting irritated with it; fix the issue once and for all, or demand a refund and move to another provider. Sep 2 '11 at 15:04
  • i use approximately 7 stackexchange sites, this one has the least polite people. welcome to serverfault! Jun 20 '13 at 21:45
5

There is only one answer to this. Move to another provider. You are wasting your time "figuring out" things. Your time is too valuable.

If a vendor screws up once and gives you bad hardware you complain and if they handle it well (quick and efficient, simple apology and assurance it is unusual, off-set billing to make it up, etc.) then that is a reason to stay with them.

If the exact same failure happens twice, then you just leave. It isn't worth your time and money to fix them, figure them out, or determine what else to do. Just find a new vendor with a good reputation and equivalent services and let these people know you want out of your contract because of repeated hardware outages.

If you have root access to the machine there are a number of system information packages that can often determine the memory type and brand but none are particularly reliable. Unless you are running high-end hardware (HP, IBM, etc.) the base memory architecture on vanilla machines doesn't really expose a lot of meta-data about the hardware.

aside: also, as the down voting and comments show, this question would be more appropriate on webhosting discussion forums. This is a site for systems administrators dealing with technical problems. If you lack physical access or root access to your machines we're unlikely to be able to help you much. For lower-level tech questions the sister sites https://superuser.com/ or https://unix.stackexchange.com/ might be more helpful.

0
1

Have you got physical access to the machine ? Box Manufacturer ? Operating system ?

No, faulty memory is not usual IMHO.

If using linux you can use memtest86 to find out any memory errors if you suspect that there could be any.

2
  • By access you mean physical access, out of band access (IPMI, LOM) or ssh access? Sep 2 '11 at 13:51
  • If it is a managed box, then you can do nothing about hardware errors. But hosting companies usually use good hardware, their business depends on it after all (and usually yours!).
    – drcelus
    Sep 2 '11 at 14:15
0

This is actually a common problem with mismatched or faulty memory modules. EDAC is a subsystem in the Linux kernel that detects and handles such memory errors, more documentation is available in the kernel doc/edac.txt file.

Not the answer you're looking for? Browse other questions tagged or ask your own question.