topic Re: pv time out in LINUX in Operating System - Linux

pv time out in LINUX

skt_skt — Wed, 13 May 2009 14:20:44 GMT

RHEL AS servers (bothe 32bit and 64 bit)
LVM1 and LVM2 version

I am looking for an equivalent in LINUX for "pvhchange -t xx" which is common in HP-UX.

I never changed the default settings in LINUX. So is there some thing at the LINUX kernel which takes care if there is a "time out" scenario.

How its done at your shops.

Re: pv time out in LINUX

Steven E. Protter — Wed, 13 May 2009 14:41:17 GMT

Shalom,

lvm for Linux was supposed to work like HP-UX LVM.

I think the man page for pvchange should have your answer.

Changing time time out rarely solves this issue which normally comes from disks that are too busy.

SEP

Re: pv time out in LINUX

skt_skt — Wed, 13 May 2009 15:09:02 GMT

I had already checked the man b4 posting here. I dont any details of time out setting there

Re: pv time out in LINUX

Heironimus — Wed, 13 May 2009 21:53:36 GMT

You won't find it in Linux LVM because there is no 1:1 mapping of LVM PVs to disk devices in Linux - it's quite common for a PV to be a partition, which isn't permitted on HP-UX. You'll need to control it at the block device or HBA layers.

I'm not sure if it's an exact match for the HP LVM option, but you have some control over timeouts on 2.6-based systems through /sys/block/DEVICE/device/timeout, where DEVICE is a whole-disk device name like sda.

Re: pv time out in LINUX

skt_skt — Wed, 13 May 2009 23:32:14 GMT

which LINUX version u r refrring about /sys/block/DEVICE/device/timeout?

Re: pv time out in LINUX

Heironimus — Fri, 15 May 2009 13:53:04 GMT

I think it should be available on most 2.6 kernels (2.4 and earlier kernels don't have sysfs). I don't have many sample distributions in front of me at the moment, but at a minimum I have a /sys/block/sda/device/timeout on RHEL4 and RHEL5.

I'm not sure if there is a way to set a global default, and I don't know if there is a ready-to-use tool for managing settings under /sys like sysctl does for /proc. Calling scripts from udev would be a reasonable solution, especially for places that are already using udev for persistent device naming.