nalbada
Contributor
Contributor

looking for information on nexus memory leak

Someone told me about a "known issue" with the Cisco Nexus 1000v causing a memory leak in ESX hosts. I am guessing that they mean the VEM is the cause of the memory leak. Over time the ESX host runs out of memory and crashes. They said there is a patch to fix this. Does anyone have information about this issue and the patch for it?

Thanks,

Daniel

0 Kudos
2 Replies
lwatta
Hot Shot
Hot Shot

A memory leak could happen in certain older versions. What version of Nexus 1000V are you runnning?

Do you suspect a memory leak or just concerned because it could happen? I've only seen a few customers actaully run into the memory leak issue. It really takes a special set of circumstances to happen.

If you want to check you run the following command

On the N1KV VSM run the below command

N1KV->module vem # execute vemcmd show heap

Look to see if "denied" or "Denbytes" are greater than zero. Also look at "objbytes" and "objs" if it is constantly growing but never receding then the VEM is grabbing memory but never giving it back.

What

nalbada
Contributor
Contributor

I am an analyst at my company, and sadly do not have direct access to the equipment (despite being certified). A division of the company is experiencing the memory leak issue, and is using that as a business justification for certain problems and activities. Unfortunately, details are not very forthcoming, and there is not any way for me to get more regarding the environment having the issue. I do not know what version of the nexus 1000v they are using, and think the environment is 3.5 since upgrading the hosts to 4.0/4.1 is being presented as part of the solution.

I guess you could say that I am double-checking them and looking for specific information to take back to my management. With the battle-cry of "known issue" being thrown around by these techs, I would like documentation of some kind that demonstrates this is a known issue. They make it sound like VMware has a patch for this, but if I am understanding you right, it sounds more like the older infrastructure is the issue.

0 Kudos