1 Rookie
•
35 Posts
0
1709
July 28th, 2021 12:00
Dell iDRAC vs ESX Memory Statistics
We recently deployed two Dell servers to serve as ESX hosts for a small office in our company. Each host has 96GB of memory. Despite the fact that they have 28 VMs, they use a very minimal amount of memory collectively. The Dell IDRAC memory stats show roughly 10 - 20% of memory utilization consistently. Having said that, the Host summary tab in vCenter constantly throws up "Host Memory Usage" alerts and shows that the hosts are nearly out of memory, even though iDRAC shows very little memory in use. This office wants to add a few more VMs but I was concerned about the memory allocation. After consulting with a VMware expert, we were told my analysis was correct and that since Dell showed little memory actually in use, we should be able to overcommit and add the additional VMs.
Four VMs were added and powered on. We experienced an outage in which the hosts froze up and VMs were inaccessible until we powered the new VMs off. During a second attempt in which they were powered on one at a time, a less drastic but similar problem occurred after powering on the second VM.
To provide a fuller picture, two of the VMs are NetApp OnTAP Select nodes, a software-defined storage solution. The datastores the VMs are hosted on reside within those VMs.
We have cases open with VMware and NetApp but are having trouble making headway so far. Does anyone have insight into this issue? I can't understand why we would need more physical memory, yet the symptoms seem to indicate we do. The only VMs with reservations are the OnTAP Select nodes, with a 16GB reservation on each.
DELL-Young E
Moderator
•
5.2K Posts
0
July 28th, 2021 20:00
Hi, hope this helps:
"
Memory Utilization — RMCs measure memory traffic occurring at each memory channel or memory controller instance. Data from these RMCs is aggregated to measure the cumulative memory traffic across all the memory channels on the system. This is a measure of memory bandwidth consumption and not amount of memory utilization. iDRAC aggregates it for one minute, so it may or may not match the memory utilization that other OS tools, such as top in Linux, show. Memory bandwidth utilization that the iDRAC shows is an indication of whether workload is memory intensive or not."
Tim_1965
1 Rookie
•
35 Posts
0
July 28th, 2021 12:00
To add one comment: I notice Dell says this about the memory stats: "This is a measure of memory bandwidth consumption and not amount of memory utilization. Memory bandwidth utilization that the iDRAC shows is an indication of whether workload is memory intensive or not." (link below) Not sure precisely what this means but it makes me wonder if we are misinterpreting the iDRAC stats.
https://www.dell.com/support/manuals/en-us/idrac9-lifecycle-controller-v3.3-series/idrac_3.30.30.30_ug/monitoring-performance-index-of-cpu-memory-and-input-output-modules?guid=guid-61137907-32af-4f1f-a0f4-e11954e1196a
Tim_1965
1 Rookie
•
35 Posts
0
July 29th, 2021 08:00
Thank you @DELL-Young E ! That's it - it was a fundamental misunderstanding of the stats being presented in Dell's iDRAC. I opened a case with Dell this morning and they clarified that the stats show traffic between CPU and memory, not memory utilization, which is basically what the article was saying that you quoted. So, we simply have a memory allocation issue on the hosts we need to resolve. Thanks!