so the name is a little misleading, the nmddtrace collects statisctics which are fed into the database for analytics, the gui reads from that database so kill the service will make the gui useless.

So this is something we are seeing a bug filed against.

In the meantime, it is possible that multi-nms is complicating the situation. Can we suggest disabling multi-nms to see if that helps?

to do that inside nmc

setup appliance nms property srvpoolcntinitial -p 0

also you didn't say how much ram you are using?? also dedup can cause alot of cpu util.

The VM has 16GB assigned to it.

Dedup is disabled and was never enabled.

Even after setting the "setup appliance nms property srvpoolcntinitial" to 0 the utilization still shows steady at 50%.

This is with no data in the pool (single pool at this point) and no NFS traffic.

Of course, prstat still shows utilization at 0-1%.

I guess our trial isnt doing so hot...

thanks for your the quick response.

-m

This is a common problem which I'm baffled as to why Nexenta hasn't fixed over the past year. I suspect this problem has extremely hurt your top line due to the fact that most people are going to test Nexenta on VMware before trying it on physical hardware to learn how the product works, only to find out that it has a slow web GUI due to the high CPU utilization and slurps up so much CPU that people believe that the same will occur on physical hardware.

Running on an 8 vCPU VM results in 50% CPU utilization, similar to smaller-vCPU VMs. Something is obviously not working as expected for each CPU.

I'm very interested in a solution as well and hope that the fix is either included in the release or the workaround is next to the download link on your site for the VMware image.

Eric