Service Console Memory (RAM)

[big_vern]

Looking for some advice on this, running HP blades with 32GB RAM, noticed the service console has only 272MB assigned. (esx 3.5)

What are the advantages to increasing this to 800MB, what, if any performance benifits will you get ??

Can the service console dynamically increase its RAM allocation if needed ?

[ctfoster]

In earlier versions of ESX Server, the service console required a specific amount of memory overhead for each virtual machine running on a host.

However since ESX Server 3.0 individual VM process threads are handled directly by the VMkernel, thereby eliminating the need to allocate service console memory.

The memory should be increased if you are planning to run apps other than the default install. This includes management agents such as HP and Dell.

Also thats a no to the dynamic allocation question.

[thomps01]

If you're planning to load any agents inside the COS, you should increase the memory.

Also, the number of running VM's dictates what you should set this to.

256MB for less than 10 vms

512MB for 10 - 19 vm's

800MB for 20+

I normall stick with 800MB

[big_vern]

thanks for the reply, are those figures documented anywhere - or is it from experience ?

[thomps01]

I did take this from an official document, but actually the other guy might be correct in saying these values were only applicable to ESX 2.5 and VI3 uses the RAM from VMkernel.

[Yattong · Accepted solution]

Actually had it confirmed with vmware support that 272MB memory is really only for a single esx host with a few vms to mess about with.

Adding a esx host to a VC, enabling HA, DRS and all the other gubbings means it needs more memory.

Attaching to SAN? HP/Dell snmp agents? All adds up.

Even though the vmkernel is not directly linked to the vm memory usage as mentioned above, it still does needs to increase as the number of vms increase.

Why? Because, the SC still needs to know whats going on with the vm's, heartbeats, HA, and further more, if its in a cluster it needs to know about all the vm's in the cluster, not just the ones on itself.

I recommend maxing it out to 800MB.

Good Luck

[big_vern]

thanks for the quick replies, 800MB is the way I'm going !

[petedr]

We have been recommended at least 500MB or 800MB if you can. One main benefit is if you are doing alot of snapshot activity either for your own purposes or from backup software.

[IB_IT]

echo-ing what has already been said. I'm not sure why the default is 272...In my experience it almost seems like they took the minimum recommended memory amount and made it the default. I have several 2-node clusters...we are set at 500 MB console memory for each host. Only application on the hosts are OpenManage.