RE: RE: [Opendlm-devel] ODLM/OGFS Recovery

Daniel McNeil <[email protected]> 26 Apr 2004 16:23:08 -0700
Newsgroups gmane.comp.file-systems.opengfs.devel
Message-ID <[email protected]>
This DLM/recovery discussion has been a very interesting.  I just
thought I would add my 2 cents worth from previous experience with
a cluster file system from several years ago.  This is from my
memory, so this might not be exact.

We had to work out these same issues.  We also used a DLM that provided
persistent locks (obviously not open DLM).  The way the DLM worked was
that after a crash (node crash), persistent locks for a particular
lock domain would continue to return DUBIOUS even after DLM recovery
until the application made a call to re-validate all persistent locks
for the domain.

The cluster file system then used deadman locks (not persistent) to 
monitor the health of the other nodes that had a particular file system
mounted (we used 1 domain for each file system).

When a system crashed:

1. The DLM lock recovery would clean up the locks from the dead node,
   but new persistent locks and converts on existing locks would return
   DUBIOUS.  (It might have have only been existing locks who were held
   by the dead node -- the surviving nodes just had them opened)

2. The deadman locks would covert since they are not persistent.

3. The surviving nodes would then decide who should run file system log
   replay for the dead node(s)

4. One node would run log replay while the other nodes waited.

5. After log replay, one node would make the DLM call to clean up the
   persistent locks.

6. Continue on as normal.

Some of these steps were a bit more complicated than I have stated,
but basically this is how things were handled.

I'm not sure how this helps or not, but the description of how
persistent locks work with open DLM seems difficult to use.  The
whole point of persistent locks was to get a notification that
a resource protected by a lock might need cleaning up without having
to keep the lock open all the time.

Daniel




-------------------------------------------------------
This SF.net email is sponsored by: The Robotic Monkeys at ThinkGeek
For a limited time only, get FREE Ground shipping on all orders of $35
or more. Hurry up and shop folks, this offer expires April 30th!
http://www.thinkgeek.com/freeshipping/?cpg=12297