cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
311
Views
0
Helpful
7
Replies
Saad_Ahmed
Beginner

qeury for expand cluster

Hi Guys 

I have an HX cluster consists  of 4 nodes 240c M4SX and 2 FI 6200 series 

The cluster is fully utilized compute so I need to expand more compute by adding a compute-only node and there are 10 desktops that require GPU but the cluster doesn't have any GPU Cards.

 

Queries?

 

1- Can I add one compute-only node M5 with different Compute resources from existing nodes in a cluster ? or must be the same compute resources?

 

2- For VMs that require GPU card, Can I add a GPU card to a compute-only node? or must add it to all nodes in the cluster?

 

 

 

 

 

7 REPLIES 7
Kirk J
Cisco Employee

Greetings.

While I believe you can add the compute node to the cluster with a GPU, it will cause other complications in regards to the automation of upgrades.  As the vGPU enabled guestVMs won't be able to vmotion anywhere, those will always need to be manually powered off, when you are upgrading UCS or ESXi.

Also, don't forget to enable EVC mode on your Existing cluster prior to trying to expand.

https://kb.vmware.com/s/article/1003212 

 

Kirk...

thanks, Kirk 

 

But what about this query

Q- Can I add one compute-only node M5 with different Compute resources (Ram + Processor )of existing nodes in a cluster ? or must be the same compute resources?

The EVC mode will filter the CPU differences, but HX installer should not error because of the different RAM capacity.

Equal RAM capacity is DEFINITELY a best practice.

However, in your particular scenario, the GPU has already created an 'island' effect for guestVMs on it, so as long as you have adequate RAM for the guestVMs, the RAM capacity variance shouldn't matter.

 

Kirk...

 

 

Thanks, Kirk
I appreciate your efforts

I have another inquiry 

I plan to scale up my HX Nodes  (HX 240c M4) by adding more RAMs in available empty slots, but I faced a problem with memory speed.

The memory which was installed 8 x 32GB DDR4-2400-MHz RDIMM/PC4-19200/dual rank/x4/1.2v and I plan to add 4 x 32GB DDR4-2666-MHz RDIMM/PC4-21300/dual rank/x4/1.2v.

That because the memory speed 2400 MHz has been ended of the sale by the vendor and the available memory speed now to purchase is 2666 MHz, but I'm not sure does it compatible with HX240c M4 or not?
Does memory speed 2666 MHz work with HX 240c M4?

xr-rules
Enthusiast

Hi,

 

you may mix that with compute but also converged nodes (disk requirement still applies until heterogeneous drive support is out, but this will be restrictive as well).

 

 

Thanks all 
I appreciate your efforts

I have another inquiry 

I plan to scale up my HX Nodes  (HX 240c M4) by adding more RAMs in available empty slots, but I faced a problem with memory speed.

The memory which was installed 8 x 32GB DDR4-2400-MHz RDIMM/PC4-19200/dual rank/x4/1.2v and I plan to add 4 x 32GB DDR4-2666-MHz RDIMM/PC4-21300/dual rank/x4/1.2v.

That because the memory speed 2400 MHz has been ended of the sale by the vendor and the available memory speed now to purchase is 2666 MHz, but I'm not sure does it compatible with HX240c M4 or not?
Does memory speed 2666 MHz work with HX 240c M4?

Officially only what is on the spec sheet is "supported" (been tested/QA'd/guaranteed to work):

  https://www.cisco.com/c/dam/en/us/products/collateral/servers-unified-computing/ucs-c-series-rack-servers/c240m4-sff-spec-sheet.pdf

 

Unofficially DIMMs typically "slow down" to match the highest common speed of DIMMs, other DIMMs, CPU, DIMMs-per-channel, etc.

The 2666 MHz DIMMs should operate just fine at 2400 MHz, or slower.

Verify the new DIMMs don't break any of the several gotchas listed on spec sheet (like don't mix RDIMM and LRDIMM).

 

Basically: Should work, can't guarantee it will work, your mileage may vary.

 

Hope that helps.

Content for Community-Ad