Im running VCOP 5.8.1.0
I have a linux vm that VCOPS report having 0 days of "Memory Effective Demand"
HEre is the memory information from whithin the VM
free
total used free shared buffers cached
Mem: 1020128 859212 160916 0 23656 253160
-/+ buffers/cache: 582396 437732
Swap: 524280 0 524280
From my understanding of Linux memory usage, there is nothing wrong with this VM,
* It even have ~160MB mem free
* And with ~250MB of cached memory that can be freed if a stress occurs in a process
To me there is no way that this VM have memory usage issue, and I dont understand why VCOPs flag this VM with memory issue. To me the memory calculation is wrong for Linux VM and vmware, they do not take into account the cache memory that will be freed by the kernel.
Can VCOPs takes into consideration the cache memory of a VM when calculating the "free" memory?
I also see this bad (to me) reporting behaviour in vcenter about the memory usage of this VM and mostly all linux VM that have a decent memory workload.
See this images from Vcenter about this VM
two key points here
- vcops does not work with the VM's internal counters, as those are misleading in a virtual environment.
The VM itself does not have a clue what's going on with the ESX, whether it's swapping or not etc. I'm not the expert in VI administration, but basically you *can't* tackle it like a physical environment.
- Time Remaining is not based on the immediate/current value - e.g. it does not represent the instant memory/disk/cpu consumption that you are having _right_now_.
For that is the Workload badge. Instead, it's telling you that there have been a steady trend of increase of the memory consumption and overall, your VM is experiencing memory deficiencies.
-Alex
Thanks for the answer.
What curious is that I have a yearly graph of this VM, and there is no ramp up of memory usage.
Maybe the VCOPs dont have enough data to properly guess the memory growth. I have the VCOPs running for one week now.
I am just trying to understand the various alerts it give me .
vC Ops capacity planning in v 5.8.1 doesn't consider the GuestOS stats of a VM. All it cares about is what the virtual hardware sees, or what vCenter/vSphere sees.
Do a graph in vC Ops of "Memory|Demand(KB)", view the last ~30 days of data, and you'll see roughly the trend vC Ops is using.
The time remaining is also based on your config policy.. e.g. what capacity planning criteria you're using. If you don't believe vC Ops's 0-days remaining, then you should revisit this policy and do a reality check on what criteria you're using for cap planning/usable cap, etc.
Thank you both of you, that make sense if VCOPs only look at the reported information. Thanks.