VMware Cloud Community
CiscoKid
Enthusiast
Enthusiast

Maximum hosts per HA/DRS Cluster recommendations

We are currently managing 2 separate instances (Development and Production) of vCenter as illustrated below:

Production vCenter Manages:

  • PROD1_CLUS - 8-HP Proliant BL480c Servers (spread among 3 chassis) - Storage Volumes=EVA6400A-PROD1-VMFS1-8 (8 paths), EVA6400B-PROD1-VMFS1-8 (8 paths)

  • PROD2_CLUS - 8-HP Proliant BL480c Servers (spread among 3 chassis) - Storage Volumes=EVA6400A-PROD2-VMFS1-8 (8 paths), EVA6400B-PROD2-VMFS1-8 (8 paths)

  • DMZ_CLUS - 4-HP Proliant BL480c Servers (spread among 2 chassis) - Storage Volumes=EVA6400A-DMZ-VMFS1-8 (8 paths), EVA6400B-DMZ-VMFS1-8 (8 paths)

Development vCenter Manages:

  • DEV1_CLUS - 8-HP Proliant BL480c Servers - (spread among 2 chassis) -Storage Volumes=EVA6000-DEV1-VMFS1-8 (8 paths), CX300-DEV1-VMFS1-8 (8 paths)

  • DEV2_CLUS - 8-HP Proliant BL480c Servers - (spread among 2 chassis) - Storage Volumes=EVA6000-DEV2-VMFS1-8 (8 paths), CX300-DEV2-VMFS1-8 (8 paths)

  • DMZ_CLUS - 4-HP Proliant BL480c Servers - (spread among 2 chassis) - Storage Volumes=EVA6000-DMZ-VMFS1-4 (8 paths), CX300-DMZ-VMFS1-4 (8 paths)

We would like to make more efficient use of our vCenter implementations and propose a new design keeping in mind that the maximum HA/DRS cluster size for both vSphere and VI3.x is 32 hosts. The change would look as follows:

Production vCenter Manages:

  • PROD1_CLUS - 16-HP Proliant BL480c Servers (spread among 3 chassis) - Storage Volumes=EVA6400A-PROD1-VMFS1-16 (8 paths), EVA6400B-PROD1-VMFS1-16 (8 paths)

  • DMZ_CLUS - 4-HP Proliant BL480c Servers (spread among 2 chassis) - Storage Volumes=EVA6400A-DMZ-VMFS1-8 (8 paths), EVA6400B-DMZ-VMFS1-8 (8 paths)

Development vCenter Manages (linked to Production vCenter):

  • DEV1_CLUS - 16-HP Proliant BL480c Servers - (spread among 2 chassis) -Storage Volumes=EVA6000-DEV1-VMFS1-16 (8 paths), CX300-DEV1-VMFS1-16 (8 paths)

  • DMZ_CLUS - 4-HP Proliant BL480c Servers - (spread among 2 chassis) - Storage Volumes=EVA6000-DMZ-VMFS1-4 (8 paths), CX300-DMZ-VMFS1-4 (8 paths)

Does the number of Paths per LUN per Host per Cluster in the above configuration cause for any concern? I know that we need to be concerned with the number of Paths per LUNs is a factor as the maximum is 1024, which means that the maximum number LUNs with 8 paths to a single host would be 128 LUNs with 8 paths. Any other considerations I have missed with the proposed redesign? What have others done with their HA/DRS cluster sizes?

Reply
0 Kudos
4 Replies
dominic7
Virtuoso
Virtuoso

I think it depends on how many blades you can get in a chassis. In your proposed plan, you have ( prod ) splitting the 16 node across 3 chassis. If these are BL460G1's that would mean you have 4 nodes on chassis1, 4 nodes on chassis2 and 8? nodes on chassis3?

If that's the case, I wouldn't do it, because losing 8 nodes ( 50% of your cluster ) would be pretty critical if you had the chassis fail. The design you have now looks to be pretty decent. If you moved to the newer BL460 G6 chassis you can get 16 blades per chassis, in that case you can spread across 3 chassis to mitigate chassis failure.

Reply
0 Kudos
CiscoKid
Enthusiast
Enthusiast

I should have been more specific with the chassis split:

Production:

  • Chassis 1: PROD1_CLUS=3-Nodes, PROD2_CLUS=2-Nodes and DMZ=1-Node

  • Chassis 2: PROD1_CLUS=2-Nodes, PROD2_CLUS=2-Nodes and DMZ=2-Node

  • Chassis 3: PROD1_CLUS=3-Nodes, PROD2_CLUS=3-Nodes and DMZ=1-Node

Development:

  • Chassis 4: DEV1_CLUS=3-Nodes, DEV2_CLUS=2-Nodes and DMZ=1-Node

  • Chassis 5: DEV1_CLUS=2-Nodes, DEV2_CLUS=2-Nodes and DMZ=2-Node

  • Chassis 6: DEV1_CLUS=3-Nodes, DEV2_CLUS=3-Nodes and DMZ=1-Node

Reply
0 Kudos
dominic7
Virtuoso
Virtuoso

So how do you plan to spread the 16 nodes across 3 chassis for production?

edit: n/n, it seems I have a reading comprehension problem.

Reply
0 Kudos
AntonVZhbankov
Immortal
Immortal

Take into consideration:

1) It is strongly recommended to have not more than 4 nodes of HA cluster in single enclosure because of HA architecture - to prevent situation when all 5 primary HA nodes will be down.

2) Single host in HA cluster can handle up to 100 VMs if cluster is 8 hosts or less, and 40 VMs if bigger.

IMHO you have now optimal configuration.


---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru

EMCCAe, HPE ASE, MCITP: SA+VA, VCP 3/4/5, VMware vExpert XO (14 stars)
VMUG Russia Leader
http://t.me/beerpanda
Reply
0 Kudos