VMware Cloud Community
Matrix_1970
Enthusiast
Enthusiast

JVM heap size and vCOPS

Hi all,

Some of my colleagues have asked me, based on their experiences as it is possible that vCOPS advised to decrease the RAM of a VM without taking into account the heap size of the JVM that is configured inside the VM for an application.

In essence, it seems that if the heap size is configured to, for example, 10GB, but I reduce the RAM of the VM from 20GB to 5GB (for example), the JVM crashes (I think the error is OutOfMemory).

In essence, then, where is the metric that include this value (from vCOPS seen that the metrics on memory do not seem to take this into account (it is as if this portion of RAM "dedicated" to the heap size is invisible when not in use).)


Thank you very much.


Matrix

Reply
0 Kudos
5 Replies
mark_j
Virtuoso
Virtuoso

The out-of-the-box oversized VM report does not take in to account application processes on VMs. This includes configured JVM heap sizes. After all, how could it possibly know this without having access to GuestOS/App-level perf counters? As of vC Ops v5.8, to expose application level performance metric to vC Ops, a different product must be integrated with vC Ops, such as Hyperic, SCOM, Introscope, ITM, etc. Even then, the OOTB reports that are run in the Advanced UI cannot be modified to include data from 3rd party adapter such as aforementioned. The OOTB oversized VM reports will judge a VM by active/allocated resources as per the configuration policy settings, but all is from a virtual hardware (vCenter stats) perspective rather than a GuestOS/application.

If you find this or any other answer useful please mark the answer as correct or helpful.
Matrix_1970
Enthusiast
Enthusiast

Thank you very much. I was hoping there was another way ...


Matrix

Reply
0 Kudos
achvn
Enthusiast
Enthusiast

Hi,

I'm OO0 with limited access to my emails until the 3rd February.

For urgent requests, you can reach me on my mobile.

Thank you,

Best Regards,

Alexis

Reply
0 Kudos
mlebied
Enthusiast
Enthusiast

Mark.j – is it feasible to write a custom report or custom dashboard to include the vcOps metric and 3rd party (Hyperic) metrics to make the oversized VM report more meaningful?

If so, what would you recommend in terms of approach

Reply
0 Kudos
mark_j
Virtuoso
Virtuoso

The oversized VM recommendations are based on demand/utilization levels based on periods and peaks defined in the configuration policy. The specific metrics used by those policies are not configurable. Let's say we were to add in a 3rd party data source, be it Hyperic or otherwise. We'd have that additional application/JVM information, but we wouldn't be able analyze it based on the periods and peaks. So you've got the intelligent information created by the configuration policy on the vCenter stats, but you'd have unintelligent(not analyzed) information from the 3rd party adapter. You can't add 3rd party data sources to the oversized VM analysis, so if you wanted to arrive at a combined "recommended VM size" you'd need to perform a little bit of human intervention to interpret the data via graphs and observing the 3rd party data to arrive at a different perspective of 'waste' within the JVM heap using garbage collection or otherwise (possibly using a Supermetric).

So as you can see, there's no easy answer right now. However, if you want to accomplish the analysis it is possible in some forms. We all look forward to this changing in future version with respect to integrating 3rd party data in to the analysis tools offered in the Advanced UI.

If you find this or any other answer useful please mark the answer as correct or helpful.
Reply
0 Kudos