Cluster - Aliasing Problems.

From: ^IT_MPG_UNIX_ADMIN <^IT__MPG__UNIX__ADMIN_at_ccmail.allegiance.net>
Date: Thu, 03 Jan 2002 10:58:27 -0600

     Would like a little bit of help understanding the Cluster Aliasing
     subsystem in V5.1. We are having some problems and I fear they might
     be fallout from incorrect implementation on our part.
     
     We have a 6 node V5.1 Cluster. 2 of the nodes are the larger size
     systems(Have more system resources) and hence we run our Database on
     one of these nodes(Node A) and the SAP Central instance on the other
     node(Node B). There is also a rather large filesystem that is hosted
     on Node B as well. All three are CAA resources and we have cluster
     aliases associated with these. In the /etc/clu_alias.config file for
     Node A & B we have the following entries -
     
     /usr/sbin/cluamgr -a selw=3,selp=1,join,alias=DEFAULTALIAS
     /usr/sbin/cluamgr -a alias=p01ap01,join,rpri=2,selw=0
     /usr/sbin/cluamgr -a alias=p01db01,join,rpri=2,selw=0
     /usr/sbin/cluamgr -a alias=prdfs01,join,rpri=2,selw=0
     
     The first being the default Cluster Alias and the rest being the three
     resouces. When the caa scripts kick in, they change the "selw" for the
     respective cluster alias to 1 on that particular node. As a result the
     system ends up hosting the service/resource/alias. The other 4 nodes
     in the cluster are never going to host either one of these resources
     and hence the only entry that they have in their /etc/clu_alias.config
     file is the one for the default cluster alias which is put in there by
     the clu_add_member script.
     
     We did have problems with the cluster aliases not responding and as a
     result applied the "Network - Cluster Aliasing" patches which have
     been since included in PK 4. But about a weekend ago, I had one of the
     other nodes crash and ended up having aliasing issues again. I could
     ping all three aliases from this particular system and could get "rsh"
     commands to work with prdfs01 but the other two aliases - "p01ap01 &
     p01db01" would not respond. I got around this problem by making this
     system join the two aliases with a selw=0 using the cluamgr command.
     Have we implemented this incorrectly somehow? Do I need to have all
     three aliases defined in each systems /etc/clu_alias.config file
     regardless? What's the supported/recommended thing to do?
     
     
     Tks & Rgds,
     
     Alay Shah
     Allegiance Healthcare Corporation
     1400 Waukegan Rd.
     McGaw Park, IL 60085
     Ph - 847-578-2584
     Fax- 847-578-5586
     Email - shahal_at_allegiance.net
Received on Thu Jan 03 2002 - 17:00:18 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:43 NZDT