I have set up 1 netRAIN interface per node on an ES40 cluster running 4.0F
patch kit 4 using DE600-AA cards with the DE600 drivers installed. The setup
of the interface via /etc/inet.local works fine and I can then start
accessing the box through the interface. A look at ifconfig returns the
following info:
ee0: flags=c63<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST,SIMPLEX>
NetRAIN Virtual Interface: nr0
NetRAIN Attached Interfaces: ( ee0 ee1 ) Active Interface: ( ee0 )
ee1: flags=c63<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST,SIMPLEX>
NetRAIN Virtual Interface: nr0
NetRAIN Attached Interfaces: ( ee0 ee1 ) Active Interface: ( ee0 )
nr0: flags=c63<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST,SIMPLEX>
NetRAIN Attached Interfaces: ( ee0 ee1 ) Active Interface: ( ee0 )
inet 10.33.128.38 netmask fffff000 broadcast 10.33.143.255 ipmtu 1500
If I then attempt to switch the interface to the backup using ifconfig nr0
switch I get the following:
ifconfig: ioctl (SIOCIFSWITCH): Invalid argument
trutest2 vmunix: NetRAIN: switching from interface ee0 to interface ee1
And the interface never actually switches. If I repeat the command it just
then returns the error message and nothing more. I also get the same result
if I physically remove the cable. This is very strange as the other cluster
member is set up exactly the same and it switches interfaces fine. What's
even more strange is that if I remove the netRAIN and add it back in, but
giving the interfaces in the reverse order, i.e. "ee1,ee0" instead of
"ee0,ee1", it then switches over with no problems. And if I do the same to
the other cluster member it then starts to get this error. The only
difference I can see is on the MAC addresses:
Cluster member where order "ee0,ee1" fails but "ee1,ee0" works
ee0 1500 <Link> 02:50:8b:ae:d0:5d
ee0 1500 DLI none
ee1 1500 <Link> 00:50:8b:ae:f4:fc
ee1 1500 DLI none
Cluster member where order "ee0,ee1" works but "ee1,ee0" fails
ee0 1500 <Link> 02:50:8b:ae:f5:67
ee0 1500 DLI none
ee1 1500 <Link> 00:50:8b:ae:e8:fd
ee1 1500 DLI none
This seems a bit weak that the order of the MAC addresses could have a
bearing on whether the interfaces switch or not but it's the only difference
I can find. I'm gonna physically swap the interfaces round on Monday and see
what happens to prove/disprove this theory. In the meantime if anyone has
had any experience of this type of problem I'd appreciate them letting me
know.
Thanks in advance,
Dave.
(dave.campbell_at_vf.vodafone.co.uk)
Received on Fri Feb 23 2001 - 17:34:06 NZDT