Hi,
We have 2 nodes on TruCluster on 5.1B. I want to install software into 1 of the node. It will install into:
i) /usr/opt
ii) /var/opt
Seems both the above mountpoints are shared.
How can I install into node1 only & make it not visible in node2. Is it possible?
Note: If you are the author of this question and wish to assign points to any of the answers, please login first.For more information on assigning points ,click
here
The cluster mechanism is designed to share resources. As result even local (not shared) disks or tapes for node1 will be "visible" to node2!
So the question as put should get the answer "not possible".
But why do you want it "not visible"? There are many ways to make it not able to start if someone tries this on node2. eg. modify startup-script If it is a cluster-resource you can configure CAA not to start the resource on node2 unless manually commanded.
>The cluster mechanism is designed to share >resources. >As result even local (not shared) disks or >tapes for node1 will be "visible" to node2!
I don't agree. I've used the following cluster solutions: 1) HP's MC ServiceGuard 2) IBM's HACMP 3) Sun's Cluster 4) Veritas Cluster
All of them have local disks ( /,/var,/opt,/usr etc) & they only have shared disks/mountpoints for DB or applications. Installing apps to /var/, /usr is not a problem, since they are local.
Depends on how you look at it. DEC "invented" clustering, so some would say that the way TruCluster does it, is the way others should...
Anyway, that's largely irrelevant. Going back to the original problem I don't see why it matters whether the second node would be able to see software installed on the first... Just because a software installation is visible to the node, it doesn't mean it has to do anything with it...
Rob, Thanks for the reply. OK, I think it is best I explain what I want to install. It is HP Operations Manager (OM8) agents. We are suppose to install it into the local disk & monitor each individual server separately. Can you advise how we can go about it in TruCluster shared (/var/opt & /usr/opt) mount points? How can /usr/opt/OV local to each node?
Hi Rob, Thanks for pointing that out. I think I'll try it out & let you know how it goes.
From the "man" page: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To copy the existing file or directory /usr/testfile2 to all members' /usr/cluster/members/{memb} directories and replace it with a CDSL that points to cluster/members/{memb}/testfile2:
# mkcdsl -a /usr/testfile2
All members have a copy of the original /usr/testfile2 file. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I wonder what's the meaning of "All members have a copy.." I just hope if I run mkcdsl in node1, node2 doesn't have it...well will try it out when I'm back to office.
Hi Rob, No good. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Node1 1) mkdir /usr/opt/OV
Node2 1) Automatically /usr/opt/OV exists
Back to Node1 1) mkcdsl -a /usr/opt/OV 2) ls -al /usr/opt/OV, it points to ../cluster/members/{membr}/opt/OV
Node2 1) /usr/opt/OV still exists & points to the same ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Conclusion = didn't achieve my objective of /usr/opt/OV only exist in Node1.
Thanks John, I've logged a call to the HPOV support. The engineer still finding info on how to install in TruCluster. So, in the mean time, instead of idling, I try asking here.
Hi Rob, Thanks for showing in details your test results. If you checked my post, I didn't create any file after creating the /usr/opt/OV & running the mkcdsl -a in node1. I'll test again & let you know the results.
Hi Jason, have been away for a few days. but I want to get back on your post on Oct 17, 2009 13:47:26 GMT ----------------------------------- Pieter, you said: >The cluster mechanism is designed to share >resources. >As result even local (not shared) disks or >tapes for node1 will be "visible" to node2!
I don't agree. I've used the following ..... So, TruCluster seems to be the "odd man" here. -----------------------------------
This is a TruCluster forum, so from my pouint of view I was referring to Tru64/TruCluster environment. where /usr and /var are cluster resources
typhoon2 (system)> mount cluster_root#root on / type advfs (rw) root2_domain#root on /cluster/members/member2/boot_partition type advfs (rw) root1_domain#root on /cluster/members/member1/boot_partition type advfs (rw) cluster_usr#usr on /usr type advfs (rw) cluster_var#var on /var type advfs (rw)
============================= now back to the original problem I don't think in our cluster environment any specific trick has been done to make the agents node-specific.
Allthough it seems not version 8 yet, I cant't imagine the installation is so much different. typhoon2 (system)> setld -i |grep -i Agent CPQIM360 installed HP Insight Management Agents V3.6 for Tru64 UNIX OPCAGT000 installed HP ITO Agent software version A.07.31
The installation directory is made member specific by using cdsl's typhoon2 (system)> ls -l /var/opt/OV lrwxrwxrwx 1 root system 32 Feb 22 2007 /var/opt/OV -> ../cluster/members/{memb}/opt/OV
So you may be looking for a problem that doesn't exist.
Pieter, Thanks for your reply. You are right in a certain way, but you really have to upgrade to OVO8 to "see" the problem. Anyways, we are running OVO7.31 like you without any problem. BTW, for OVO8, agents will not work if /proc is not mounted. Is the /proc in your clusters mounted?
Hi Jason, Yes, /proc is mounted. => /proc on /proc type procfs (rw) I,ve not listed all our mounted filesystems. /proc is crucial for all "interprocess communication" the system cannot run normal without it.
Printable version
