In our 2 node cluster of GS140s running T64 v5.1a and Oracle 9iRAC v9.0.1,
the cluster alias does not seem to be effectively distributing the connections
in a round-robin fashion among the nodes. I have several questions.
1. How can you tell how the load is being distributed? I'm basing my
perception on watching "top" and "monitor", and on output of
"clu_check_config" - observe:
----------------------------------------------------------------------
***** Running cluamgr on member one-icso
Status of Cluster Alias: dbclust0-app
:
flags: 5<ENABLED,IP_V4>
connections rcvd from net: 6
connections forwarded: 5
connections rcvd within cluster: 23
data packets received from network: 9268
data packets forwarded within cluster: 188
***** Running cluamgr on member two-ics0
Status of Cluster Alias: dbclust0-app
:
flags: 5<ENABLED,IP_V4>
connections rcvd from net: 70
connections forwarded: 5
connections rcvd within cluster: 3935
data packets received from network: 47815911
data packets forwarded within cluster: 73
Member Attributes:
memberid: 1, selw=1, selp=1, rpri=1 flags=11<JOINED,ENABLED>
memberid: 2, selw=1, selp=1, rpri=1 flags=11<JOINED,ENABLED>
----------------------------------------------------------------------
Member "two-icso" seems to be handling much more work based on
connections rcvd and packets rcvd, and this is also our perception
based on watching it.
2. I talked to a local HP UNIX person, and he recalled some issue
with Oracle 9iRAC on a cluster that might admire some fancy tuning
of the selection weight parameter to the cluster aliases to get
things to round-robin correctly - anyone heard of such an issue?
3. Any pointers to good tuning info for clusters, wrt. distributing
load among the nodes, how to use the selw/selp/etc?
Thanks in advance,
Judith Reed
Received on Wed Oct 23 2002 - 20:33:12 NZDT