[ofw] DHCP issues with 2.2

Fab Tillier ftillier at microsoft.com
Thu Mar 25 16:05:45 PDT 2010


Hi Folks,

I've been seeing fairly consistent issues with DHCP getting "hung". The scenario is this: cluster configured with DHCP server and OpenSM on head node, serving addresses to IPoIB.  If some compute nodes get rebooted a couple times, they (and other nodes) are no longer able to get DHCP addresses.  This seems to happen with both opensm.exe and opensm_3_0_0.exe, so it's not clear that this is an SM issue.  It could be an issue in IPoIB.  In any case, restarting the SM fixes things.

Still trying to narrow down what causes this, but thought I'd bring this up now in case someone else has seen it too.

-Fab



More information about the ofw mailing list