[ClusterLabs] Fwd: need some advise on pacemaker cluster setup with oracle db

Tom Naorem tom.sana at gmail.com
Thu Aug 25 10:42:20 EDT 2016

hello experts,

i need some advise in a pacemaker cluster setup.
i have configured a pacemaker cluster with 2 x bl460c gen9 servers on a
rhel 7.2 platform. cluster is configured and fencing is configured using
fence_ipmilan agent and it is working.

there are 16 luns from a 3par storage as a shared storage on both the nodes
and i can see them in the os as mpatha, mpathb ...etc on both the nodes as
it is through multipathing and i created lvms with vg and lv commands and i
created 8 lvms on first node and 8 lvms on the 2nd node as per customer
requirement and i have created 16 same mount points on both the nodes with
the same path for mount points on both the nodes for failover.
now when i run lvs command i can see the 16 lvms on both the nodes with
same names.
all the 16 lvms are ext4 filesystem as per customer requirement and i
created the first 8 filesystem cluster resources on the first node and the
remaining 8 filesystem cluster resources on the 2nd node and they are not
cloned so that means it will run on 1 node at a time so now the first node
has 2 resource group with 8 filesystem cluster resources and the 2nd node
has 2 more resource group with 8 filesystem cluster resources.
a/c to customer requirement i need to set 4 virtual ips (2 on each nodes)
so i created 4 virtual ips and i alligned these virtual ips one for each
resource group in the cluster so 4 virtuals ip for 4 resource group and 2
of them on each node. i tested the failover 2-3 times by rebooting each
node and they are failing over the cluster resources with all the data in
place on the mount points on both the nodes so i know the setup is working
fine till here.

now here comes my first doubt  >> do i need to create and run dlm and clvmd
cluster resouce on both the nodes for this setup, i know we need to run
those resources in gfs2 filesystem when cloned but i couldn't find a good
document which talks about dlm and clvmd on a ext4 filesystem which are not
cloned resource , or they are only necessary when it is a cloned filesystem
resource or when they are running on both the nodes at the sametime, also
note that i have not created lvm cluster resources for these filesystem
cluster resources or do i need to make some changes in the current setup.

my 2nd doubt >> do i really need to re-create the initramfs with dracut
command or it is not needed as the cluster volumes are just fine on all the
reboots during the failover test.

and the 3rd one >> customer need to run 4 instances of oracle db (12c) , 2
of them on each node according to the design doc. i found this article (
https://access.redhat.com/solutions/25970) but this doesn't help much. this
article talks about adding oracle resource in the pacemaker cluster but it
doesn't talk about multiple instances though. is it really possible to run
4 oracle db instances on a 2 node pacemaker ha cluster on rhel7.2 or oracle
rac is the only way to do this ?

would really appreciate any input on this ?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20160825/1abd01f4/attachment-0003.html>

More information about the Users mailing list