#2 trying take lock hangs after node down event

[Aneesh Kumar K.V]

this is what i did

two node ci-linux cluster.
node 1 and node 2 works smoothly by taking and
releasing locks.
bring node 2 down. Try to acquire the lock on node 1.
No response. Seems to be waiting for some information.
^C will end the application.

analysis.
seems to be a problem with state of node and lock
status. Attaching is the dmesg output when trying to
take the lock on node 1 afer node 2 has gone down.

Trying to acquire the lock with a new name works .

[Aneesh Kumar K.V]

dmesg output during lock operation

[Aneesh Kumar K.V]

Logged In: YES
user_id=230991

Another observation :
Bringing node 2 back makes the locking work again.

[Aneesh Kumar K.V]

Logged In: YES
user_id=230991

I am taking a lock using .
status = dlmlock_sync( LKM_EXMODE, & lockst, LKM_VALBLK,
LOCK_NAME, strlen( LOCK_NAME ), NULL, NULL );

and unlocking it using .
status = dlmunlock_sync( lockst.lockid,
NULL,
LKM_FORCE|LKM_INVVALBLK);

With single node in /proc/haDLM/rldb i get the following entry
72719 | Openlocktest | VMS | 0005 | 3

with two nodes.
72719 | Openlocktest | VMS | 03E5 | 3

The flag value changed from 0005 03E5.

When the same test case is run on the new node added i get

72719 | Openlocktest | VMS | 0005 | 2

After this when the new node that was added goes down any
lock operation on the Openlocktest hangs

-aneesh