If an ISCSI NetApp storage array is refreshed within Oracle VM Manager after a LUN is unmapped at the array, and then refreshed again after the LUN is mapped back the second refresh may hang.
This is due to the fact that the Oracle VM Server sees the LUN size as
zero and fails the multipath path to the LUN while the default
path checker for NetApp TUR
sees the path as up.
The vacillating path statuses can cause IO to the affected LUN to
hang.
Remapping LUNs is risky because it can data corruption since the targets have been switched outside of the server.
Workaround: One way to ensure
that the Oracle VM Server maintains a consistent view of the path and
avoids a possible hang scenario is to change the path checker for
NetApp in mulipath.conf
to
directio
. Note that directio
is
not considered as performant as TUR
. Running the
following command on the affected Oracle VM Server clears the hang:
# iscsiadm -m node -R
Bug 18421164