[Linux-cluster] suggestion require about fence_vmware

[jayesh.shinde]

Hi all ,

I am testing RHCS "Active - Passive " on RHEL 5.5 64 bit OS along with 
VMWARE ESX 4.0 for *mailing* application . I want suggestion and 
guidance that , what should be the best architecture and best practice 
for the below mention setup.

My architecture details are as follows  :--
------------------------------------
1) I have one VM called "node1"  under one physical VMWARE server.
2) second VM  called "node2" under second  physical VMWARE server.
3) Both the physical servers are connected to each other with switch 
along with NIC teaming  and fail over switch is also available.
4) For fencing I am using fencedevice agent as "*fence_vmware*"
5) one SAN partition with LVM configure.  User's mailbox data is inside 
this SAN partition.
6) File system is "EXT3"

My queries :---
=========
1) Is above architecture's  points from  1-6 are correct for 
Active-passive configuration with VMWARE. ?
2) While testing I observer that when I purposely stop the network 
service on active "node1" then "node2" fence the "node1" properly.
      But  "node1" getting fence    by "*poweroff*" .
    Since all service was running on "node1" and SAN partition was also 
mounted and suddenly "node1" get fence.

*So will this immediate poweroff cause the corruption of SAN ext3 file 
system and local HDD too* ?

3) If yes how to avoid this ?
4) Is it a correct way of fencing ?
5) Is this correct setup  for  production environment ?

I purposely stop network service on "node1"  because I want to test and 
know that , what will happen when network goes down on "node1" .
I also observe that the files which was open in VIM editor also got 
recover properly because of the Journaling feature of EXT3 FS.

Happy christmas.

Thanks & Regards
Jayesh Shinde


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20101229/fd93af85/attachment.htm>