Folks,
I've recently tracked down why gmond reports incorrect values for the
MTU of all the nodes on our cluster. The nodes have two NICs, but we
are only using one, which has an MTU of 5900 (long story, some time it
will be 9000). However, gmond is reporting an MTU of 1500 for all
nodes. Going through the code, it appears that the code in
gmond/machines/linux.c (and other machine types) is also considering
interfaces that are not up.
The patch below is for Ganglia 2.5.4 and seems to fix the problem in a
test case I have, and a simple work around appears to be to set the MTU
for all interfaces.
Frank
P.S. I'm not subscribed to this list, so please reply directly to me
with any followup questions.
--- gmond/machines/linux.c.orig 2003-07-16 10:38:47.000000000 +1000
+++ gmond/machines/linux.c 2003-10-26 00:38:18.000000000 +1000
@@ -826,7 +826,7 @@
unsigned int *min = (unsigned int *) arg;
/* Only consider interfaces that are up. */
- if (! entry->intf_flags & INTF_FLAG_UP)
+ if (! (entry->intf_flags & INTF_FLAG_UP))
return 0;
mtu=entry->intf_mtu;
