[Linux-cluster] Recommended fence timeouts

[Phillips, Kaerka]

Hi-

 

I was wondering best practices for a 4-node cluster using Dell R905's
and the fence_drac function.  The server which is controlling the
cluster keeps attempting to fence too fast and hanging the drac5 cards
(amongst other issues).  Each R905 has quad-quad core AMD Opterons and
128gb of ram.

 

This is the default which I'm using now:

<fence_daemon clean_start="0" post_fail_delay="0" post_join_delay="3"/>

And the openais config:

<totem consensus="200" join="60" token="10000"
token_retransmits_before_loss_const="20"/>

 

Thanks,

 

Kaerka Phillips

Unix Systems Services Section

USPTO - 3C21

571-272-6443

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20090323/d9b961e0/attachment.htm>