Best practice for EMS shared store on RedHat Cluster with GFS2 in VMWare/vSphere
Article ID: KB0080192
Updated On:
| Products | Versions |
|---|---|
| TIBCO Enterprise Message Service | 8.x |
Description
- Use a 3 node cluster. Use two nodes for EMS, and a third small node. This prevents a split brain situation of a two node cluster. A quorum disk works OK in a physical environment, but in a VM environment, we have seen issues.
- The fence used can make a difference. In a VMWare environment the vmware supplied fence works the best. It used to be the soap based fence.
- GFS2 should be on a raw device. Avoid using a VMDK for GFS2.
- Do not setup the RedHat cluster to manage EMS. This is counter-productive, and can cause issues starting/stopping EMS. Let the capabilities of EMS F/T do its job.
- DO NOT allow vMotion, Snapshotting, or DRS be used on the nodes in the RedHat cluster. We all know what effect these can have. (It can affect the performance of other virtual machines on your network during the process. network disconnect)
- Use the most recent version of EMS. Even, 8.4 over 8.3!
- Ensure that all VMWare components used are current supported versions with up-to-date drivers.
- Preallocate the data stores (file_minimum). It may improve write performance to GFS2 by ~30%.
 | ReplyForward |
Issue/Introduction
Best practice for EMS shared store on RedHat Cluster with GFS2 in VMWare/vSphere
Feedback
Yes
No
Powered by
