When viewing the performance data for a cluster, all the charts work fine, except for the memory charts. The memory charts show the utilization at anywhere from 60% to 120%... average is about 110%. This was happening on VC 2.0.1 and 2.01 patch 2.
Today I blew away the whole thing and did a fresh install of VC2.0.2, new SQL database as well. I'm seeing the exact same issue. The memory charts on each ESX host work fine, its just at the cluster level where we see this problem.
Does anyone else have this issue? or know if its fixable? I gave VMware a call, their answer? "Well, the product should get better over time." Helpful.
The graphs you see in VMware are using something similar to RRD to create. They get data every 5 minutes and compute rates of changes and thereby can have some odd results until it has enough data points to compute the proper rate. In general that is just 3 data points, and as much as 5.
But also consider your cluster. Are the systems in your cluster the same? DO they have like amounts of memory? If not, it could be looking at the lowest amount as a base unit. Split out your memory usage by host in the graphs to test this theory.
The answer you got from your VMware Support Agent is based on the fact that they believe you do not have enough data yet, you should ask for a Level 2 support person and get the real answer. For a real time graphs as long as you have the hosts in time sync with the VC server, after 15 minutes you get your first valid data point on the graph. For non-realtime graphs you should still get a valid data point after 15-20 minutes.
Thanks for your reply. All the servers in the cluster are identical, with the same specs.
Granted for today I only have a couple hours of data to see so far, but up until this morning I had about a months worth.
Like I said, all the other performance graphs look correct, its only the memory one thats showing this behaviour so I don't think its the amount oif data I have.
Can i join in on this? I have a similar situation. Cluster with 3 nodes (all Dell 2950 quad core with 32 gb memory)- at the ESX host level, memory avg (last day) % for
node A (4 VMs) is 26 %
Node B (1 VM) is 8 %
Node C (0 VMs) is 2.8 %
The cluster level graph shows an avg of 37 %
Shouldn't the cluster avg be around 12 % ?
I have a customer with this same issue. They are running VC 2.0.1. I was about to recommend going to 2.0.2, but based on the responses here that doesn't help. Unfortunate that VMware hasn't recongized it as an issue yet. Again, in my case, it is only when looking at memory usage of the cluster. CPU of the cluster appears fine, and the memory of the indvidual hosts is fine.
On the cluster level it doesn't look like the numbers are actually the average. It looks like they are the total of the averages for each server in the cluster.
I just checked mine and that is exactly what it adds up to.
I have the same issue here, with a twist. There are 3 identical hosts each with 24GB of memory.
Host 1 Average: 11.85%
Host 2 Average: 21.89%
Host 3 Average: 17.61%
Sum of Averages: 51.35%
Cluster Average: 51.24%
It looks like someone forgot to put in the divisor in the formula for the cluster chart.
The twist is that if I look at the daily chart, I get a regular "heartbeat" where the memory utilization for the cluster and each host goes to almost zero every hour on the half-hour.
Message was edited by:
Has anybody resolved this issue yet?
We have the same problem in the memory tab. We have two clusters and they both have the same problem.
Still broke in 2.5 as well, though it seems to happen less often.
I filed this on ticket #193909451 back in September 2007. They closed the ticket, and replied: "Description: memory
percentage in the performance of the graph in cluster level is showing
wrong number, it adds the hosts up but it wont divide them by 2
known issue. will be fixed later". They clearly define "later" much more loosely then would you or I.
I'm going to reopen it.