Should I power down a server to replace a disk?

[kw_1]

One of the disk in my R class server failed. The failed disk is a hot-swapable Jamaica disk and it is part of the boot volume. All of the disks on this server are fully protected with MirrorDisk/UX. The server is located at a remote site where I have operators maintaining servers. I told both my operator and the tech person not to power down the server for the repair. Simply swap in the new disk and re-sync the disk. Of cause, after I hung up the phone, they went ahead and power downed the server anyway. I was later told by the tech person it was necessary to power down the server, the disk was partially failed and he had to stop any I/O to the disk. I disagreed, I though the whole idea of mirrored disk is to let a system survive such abrupt interruption of disk I/O. Was I wrong?

[Ken Hubnik_2 · Accepted solution]

You are correct ....should be able to plug and play and it should start to sync.

[Pete Randall]

If you're wrong, then what is the point of hot swappable? I'm sincerely hoping (and believing) that you're right.


Pete

[Zeev Schultz]

If this disk has an active healthy mirror and
is hot-swappable in most cases (except some hung IO) there is no need to shutdown the computer.Boot volume you mean /stand (lvol1 in vg00)?What io can be to /stand...hmm,some kernel compilation? :)

[John Poff]

Hi,

We run some K-class boxes in our remote sites and we have the internal disks mirrored to Jamaica disks. We've replaced bad Jamaica disks before without powering down the server.

I'd fuss at your service guy and teach your remote operators to be more stubborn about powering down the system.

JP

[A. Clay Stephenson]

Let me put it this way, I haven't shutdown to install a disk in at least 5 years and in that time I've replaced dozens of disks at the very least.

Even in the case of a "flaky" disk, the procedure is to yank the disk, wait a wee bit until LVM decides that this is now a failed disk, and proceed with the replacement. Even to replace boot disks, no shutdown is needed so long as you are mirrored and everything is hot-plug.

Even on older boxes which didn't have hot-plug internal disks (e.g some D's and all K's) the answer was simply NEVER to use the internal drives - leave the bays empty - and use hot-plus for everything.

[Steven E. Protter]

You will see an lbolt in syslog and dmesg output after you swap the drive.

It will go away at next boot.

Just this week, I unplugged an entire Jamaica box that was mirroring my production system to facilitate daytime repairs.

Nobody noticed.

Late post, zero points please, I just like to blab.

SEP

[Caesar_3]

Hello!

If it's hotplug disk do you can do it on the
fly (if you not run from the disk or he is the primary swap)

Caesar

[Sanjiv Sharma_1]

Hi KW,

For a Hot Swap disk there is no need to shutdown the server.

Simply take out the defective and replace it with the new/good one and then sync the mirror.

Thanks,