E Lvm I/O Timeout Parameters; Logical Volume Timeout (Lv Timeout); Physical Volume Timeout (Pv Timeout) - HP -UX 11i Administrator's Manual

Logical volume management
Hide thumbs Also See for HP-UX 11i:
Table of Contents

Advertisement

E LVM I/O Timeout Parameters

When LVM receives an I/O to a logical volume, it converts this logical I/O to physical I/Os to one
or more physical volumes from which the logical volume is allocated. There are two LVM timeout
values which affect this operation:

Logical volume timeout (LV timeout).

Physical volume timeout (PV timeout).

Logical Volume Timeout (LV timeout)
LV timeout controls how long LVM retries a logical I/O after a recoverable physical I/O error.
LVM timeout can be configured for a specific logical volume using lvchange. See the lvchange(1M)
manpage for details. Also, note the following:
The default LV timeout is zero which means infinite timeout to retry physical I/Os with
recoverable errors.
If the LV timeout is changed, it should be kept greater than or equal to the PV timeout.
Set higher LV timeout for greater I/O resiliency. Set lower LV timeout for faster I/O failure
detection (which, in turn, facilitates faster failover).
Physical Volume Timeout (PV timeout)
PV timeout is the time budget set by LVM for each physical I/O originating from LVM. The mass
storage stack underlying LVM manages completion of the physical I/O subject to this timeout,
resulting in success or failure for the request.
WARNING!
timeout budget, there could be situations where the I/O completion exceeds this timeout. It could
be due to failures in recovery mechanism of underlying transport layers.
PV timeout can be configured for a specific physical volume using pvchange. See the
pvchange(1M) manpage for details. Also, note the following:
With 1 1i v3, native multi-pathing is enabled by default in mass storage stack. When it is
enabled:
For clarity, it is recommended that only agile DSFs and cluster DSFs should be used in
LVM when configuring or adding physical volumes to the volume group. Native
multi-pathing will still be in effect even if legacy DSFs are specified and this may cause
confusion
PV timeout is the time budget given to mass storage stack for its I/O retries on all of the
PV links to this physical volume. However upon the expiration of this PV timeout for a
particular physical I/O, the mass storage stack will return this I/O back to LVM even if
not all the PV links have been tried within the duration specified by the given PV timeout
value. For example, if PV timeout is 30 seconds and if the physical volume has 3 links,
the mass storage stack can utilize at the most 30 seconds to retry any I/O on as many
links as possible.
The PV timeout can be from 2 seconds to a maximum of 300 seconds. The default timeout,
shown by a value of "0", is 30 seconds. However, to ensure reliable handling of device
failure, the driver stack requires a PV timeout of no less than 15 seconds. Using values lower
than 15 seconds can lead to false failure detection and possibly longer delays as LVM is
immediately recovering from a failure. Ideal PV timeout value for a physical volume depends
on several factors including number of links to the physical volume, quality/speed/recovery-time
of the transport links, physical volume model/type, and so on.
While mass storage stack strives to ensure the physical I/O completion by this PV
Logical Volume Timeout (LV timeout)
161

Advertisement

Table of Contents
loading

Table of Contents