Hi,
I have a question about Production Server
Software: Unix 4.0E, TCR 1.5 Production
Server.
Hardware: Two 4100s, each has three
network interfaces:
tu0: ethernet interface to the subnet incl the
other 4100's tu0 and also the gateway and thus
to the rest of the world
tu1: ethernet interface to a crossover cable that
goes directly to the tu1 of the other 4100
mc0: memory channel to the mc of the other 4100
We use trucluster for DRD services which go
over the memory channel, and also for the NFS
services. I should say that the NFS services
are shared only by these two machines, i.e.
that each 4100 hosts three services, and all
six services are mounted on both machines, but
not on any other computer.
during busy times we are getting slow response
and a bunch of "NFS3 server xxxxx not responding
still trying" type messages on each machine from
all the NFS services mounted on the other machine
(but not from those mounted on itself), and I'm
wondering whether we have it configured right and
whether we can configure it better.
The service status for services in asemgr looks like
this for a service 'nfs01' with a mount point 'CRDB':
The two hosts are crdb1 and crdb2 (these are the
names of the tu0 interfaces), the memory channel is
mc_crdb1 and mc_crdb2 and tu1 (the directly
connected interface) on each machine is crdb1i and
crdb2i respectively.
----------------------------------------------------
Status for NFS service `nfs01`
Status: Relocate: Placement Policy: Favored Member(s):
on mc_crdb1 yes Favor Member(s) mc_crdb1,mc_crdb2
Storage configuration for NFS service `nfs01`
NFS Exports list
/CRDB -root=crdb1:crdb1i:crdb2:crdb2i -rw=crdb1:crdb1i:crdb2:crdb2i
Mount Table (device, mount point, type, options)
domain01#CRDB /var/ase/mnt/nfs01/CRDB advfs rw
Advfs Configuration
Domain: Volume(s):
domain01 /dev/rz33a
----------------------------------------------------------
My question is, I guess, in two parts:
(1) The Exports list looks weird - is it right that it shows the tu
interfaces but not the memory channel interfaces - maybe
it should be:
/CRDB -root=mc_crdb1:crdb1i:crdb1:mc_crdb2:crdb2i:crdb2
-rw=mc_crdb1:crdb1i:crdb1:mc_crdb2:crdb2i:crdb2
This cluster was once an ASE cluster (hence the second
network interface, intended exclusively for ASE traffic)
and I'm wondering whether at PS (and thus memory
channel) installation time the then system manager should
have updated the exports list and didn't!
(2) in general, do you have any suggestions on
how best to utilize the different network interfaces
for best performance? If we use the memory
channel for all TruCluster-related traffic maybe we
should make tu0/tu1 a NetRAIN or something?
Thanks, I'll summarize
John
Received on Thu Dec 16 1999 - 17:02:39 NZDT