Storport resets the lun after hitting request timeouts. However,
it never actually removes the disk. Having a stale disk can be
troublesome, leading to an unresponsive host in certain situations
(e.g. cache deadlocks, hanging persistent reservation
requests, etc).
For this reason, we'll detect stale connections and disconnect
the disk. This feature along with the timeouts are configurable.
By default, we'll consider a connection to be stale if at least one
request older than 15s got aborted and if no IO reply was received
in the last minute.
At the same time, we'll include the following timestamps in the
wnbd-client.exe stats output:
Storport resets the lun after hitting request timeouts. However, it never actually removes the disk. Having a stale disk can be troublesome, leading to an unresponsive host in certain situations (e.g. cache deadlocks, hanging persistent reservation requests, etc).
For this reason, we'll detect stale connections and disconnect the disk. This feature along with the timeouts are configurable. By default, we'll consider a connection to be stale if at least one request older than 15s got aborted and if no IO reply was received in the last minute.
At the same time, we'll include the following timestamps in the
wnbd-client.exe stats
output:Signed-off-by: Lucian Petrut lpetrut@cloudbasesolutions.com