[PVFS2-users] What is "Error: bad magic in BMI TCP message."?

[Meik Hellmund]

I am testing pvfs2 on a small cluster. Installation and first tests went fine
but now I encountered a problem. Any advice would be appreciated.

My configuration:
- 4 Dual-Opteron nodes, connected by a Gigabit network, 
  Debian's amd64 distribution, pvfs2 1.0.1 
- Every node is a pvfs2 server providing a local 50G partition so that I get a
  200G pvfs2 partition.
- On ervery node runs a client connecting itself to the local server.

What happens: from time to time, processes that access the filesystem (i.e. an
stat(2) call), hang forever. The pvfs2 server logs:

"Error: bad magic in BMI TCP message."


Another question: is it possible to use pvfs2 with the LAM implementation of
MPI via ROMIO?

Any advice would be appreciated. Thank you for your time!


Meik

# pvfs2-ping  -m /wrk

(1) Parsing tab file...

(2) Initializing system interface...

(3) Initializing each file system found in tab file: /etc/fstab...

   /wrk: Ok

(4) Searching for /wrk in pvfstab...

   PVFS2 servers: tcp://localhost:3334
   Storage name: pvfs2-fs
   Local mount point: /wrk

   meta servers:
   tcp://node0:3334

   data servers:
   tcp://node0:3334
   tcp://node1:3334
   tcp://node2:3334
   tcp://node3:3334

(5) Verifying that all servers are responding...

   meta servers:
   tcp://node0:3334 Ok

   data servers:
   tcp://node0:3334 Ok
   tcp://node1:3334 Ok
   tcp://node2:3334 Ok
   tcp://node3:3334 Ok

(6) Verifying that fsid 2135043844 is acceptable to all servers...

   Ok; all servers understand fs_id 2135043844

(7) Verifying that root handle is owned by one server...

   Root handle: 1048576
   Ok; root handle is owned by exactly one server.

=============================================================

The PVFS filesystem at /wrk appears to be correctly configured.
 

-- 
Meik Hellmund
Institut fuer Mathematik, Uni Leipzig
e-mail: Meik.Hellmund at math.uni-leipzig.de
http://www.math.uni-leipzig.de/~hellmund