Comments | geddcoon: sysadmin, another one of my thankless jobs

geddcoon

sysadmin, another one of my thankless jobs

Jan 15, 2012 20:06

One of the physical disks in a logical volume "forgot" its UUID at some point today. The thing has been running great for 4 months and now I get some weird error. No clue as to why except that I'm nearly 100% sure that it's not a "bad disk" or other hardware error since the underlying physical disks are actually virtual disk images and the ( Read more... )

Comments 9

wyldside January 16 2012, 11:21:19 UTC

What base is this system using? Linux, freebsd, windows, unix?

geddcoon January 16 2012, 17:30:52 UTC

CentOS guest on XenServer host.

wyldside January 19 2012, 10:49:01 UTC

late reply but i'd guess the xenserver updated and regenerated the uuid

geddcoon January 19 2012, 21:18:12 UTC

From what I understand, the UUID is a data structure that is written by the LVM under Linux. There's no reason for the underlying XenServer to touch it or even know about it.

Thread 5

peewee180 January 17 2012, 19:35:40 UTC

can you run a S.M.A.R.T. test on the HDD?

clipfox January 17 2012, 20:03:18 UTC

That feature is an acronymic misnomer though... Whomever came up with the current implementation(s) SHOULD'VE named it: S.T.U.P.I.D. That thing is almost never useful. And they set the warning/failure threshold so low, that the drive has to be on fire and spewing noxious smoke. It'll report right before it's about to explode...

But, it's probably not a physical problem, since the drives are in a RAID 5 container, the RAID controller hasn't generated warnings that I heard of, and other VMs are not failing.

geddcoon January 17 2012, 23:42:00 UTC

What Clip said.

The HDDs are fine, the (fancy, expensive, enterprise) RAID controller monitors its health.

re: additional layers of complexity mejeep March 3 2012, 22:40:35 UTC

My intuitive half agrees with you. As Seymour Cray used to say "it's better when it's real" (regarding virtual memory and perhaps the MMU: memory mangling unit). The beauty of Unix and Linux is how it's layered "just enough", not like the mega-layered Multics.

But I just attended an IBM presentation mostly about cloud & virtualization on their power-7 servers. Wow. IBM has hardware support for the virtual LAN and has always used intelligent controllers for I/O. That's the power of the mainframe: many co-processors and controllers working together and tightly coupled.

But cloud storage scares me because that's totally magic and "pay no attention to the man behind the curtain". It's a way to outsource your disk storage. And companies somehow buy into it since it's not their administrative headache. But how is the reliability and security REALLY achieved? Nobody's really telling. How is one to trust a service that cannot be verified or validated.