[Linux-cluster] dlm_recvd and dlm_sendd causes high load.

[Dan Deshayes]

Hello everyone,

I've setup a high availibility cluster but have some problems with 
dlm_recvd and dlm_sendd processes.
Everything seems to be running smooth except these processes that takes 
half of the cpu power each.

I'm running a 2-node cluster using Centos Beta5 with own compiled 
2.6.20.4 kernel using a shared firewiredisc for
the shared storage with GFS2 lock_dlm.

When i start the clvmd-service the messagelog says:

clustertest-2 kernel: dlm: clvmd: recover 1
clustertest-2 kernel: dlm: clvmd: add member 1
clustertest-2 kernel: dlm: clvmd: add member 2
clustertest-2 kernel: dlm: got connection from 1
clustertest-2 kernel: dlm: clvmd: total members 2 error 0
clustertest-2 kernel: dlm: clvmd: dlm_recover_directory
clustertest-2 kernel: dlm: clvmd: dlm_recover_directory 0 entries
clustertest-2 clvmd: Cluster LVM daemon started - connected to CMAN
clustertest-2 kernel: dlm: clvmd: recover 1 done: 7049 ms

After the startup of clvmd the processes tops the process list.

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 2473 root      20  -5     0    0    0 S 49.8  0.0   5:06.48 dlm_recvd
 2474 root      20  -5     0    0    0 S 49.8  0.0   5:07.40 dlm_sendd

I've been trying to find some useful information but with no success.

Anyone having an idea or experienced the same problem?
Maybe I can provide more useful information on this?

Thanks in advance,
Redgards Dan.