DRS issue in vSphere 5.0

[HemantPatel]

Hi,

Below is my enviornment.

1. 6 ESXi 5.0 Server configured with HA / DRS (Fully Automated)

2. DPM is off

3. vCenter Server 5.0

What i noticed here 2 server out of 6 is consuming more than 95% RAM still DRS is not moving VM's to my other ESX host which is havin very lower load or zero load.

Is this normal behavior or is there any thing wrong in configuration. I am attaching screen shot of my configuration for your reference. I am also attaching network settings on all my ESX host. I have to two 1 GB NIC teamed and carry vmkernal and vm traffic.

Regards,

Hemant 

[Virtualinfra]

why one of the esxi host is disconnected.

Refer DRS deep driver in the below site.For how drs works?

http://www.yellow-bricks.com/drs-deepdive/

3 parameters to note

1. Migration Threshold
2. Target host load standard deviation
3. Current host load standard deviation

for DRS to work properly its recommended to use same size of esxi with RAM and CPU.

out of 6 ESXi 2 of them are 30+GB and rest are with 60+ GB in your screen shot.

Turn OFF DRS once and TURN ON DRS on the cluster and see if that resolve the issue. this might be because of one the esxi displays as disconnected and also imporper size of resources esxi host with a cluster.

Award points to helpful and correct answer by clicking the below tab

[HemantPatel]

Hi Darshan,

Turning off and on DRS would not help me.

One ESX is down today due to some issue in OS.

Yes you are right that that is a good practice to have same resource on all ESX in a cluster but right now i am doing POC in my company to implement HA and DRS in a all group so selected any server.

regards,

Hemant

[abbie11]

Hi have you enabled the cluster ,if not Run the vCenter Server virtual machine in a
vSphere HA enabled cluster and configure the virtual machine with a vSphere HA restart priority of
high.

Yours, Abbie

[Rubeck]

Migrating a VM manually using vMotion works just fine?

/Rubeck

[weinstein5]

As long as the VMs are receiving the reqourses they require DRS will not move the VMs - you can test but placing an artificial load on the C{U of one of the VMs - and it shoudl get vmorioined to another host - 

[Rubeck]

I agree with this statement, but as vCenter sees the cluster being imbalanced an administrator might be wondering why in the world it dosn't trigger a vMotion of a few VMs..? Triggering alarms and warnings is easier, I guess 🙂

DRS is at times stupid.. You might also expect a VM to powered on at the host with 0 VMs already running in a cluster... but even this dosn't seem to be the case at times.. 

The Q of moving the VM manually was actually due to I can't see if vMotion is enabled on the vmKernel interface...

/Rubeck

[weinstein5]

You would not have been able to enable DRS if a vmkernel port was not enabled for vmotion since DRS relies on vmotion to function - 

[Rubeck]

To this I have to disagree, as I've just done it...

/Rubeck

[weinstein5]

Hmmm - intersting - I thought DRS confirmed a check that vmotion is enabled before you can enable - you learn something new every day - but without vmotion ports then DRS will not be able to mograte a VM if it is not able to get its required resources - 

[Rubeck]

Yes, logical thinking would assume this...

That's one of the reasons I stated that DRS is stupid at times as it seems that vCenter does not check host config requirements at all..

Im sure that you can create DRS enabled clusters without hosts and then add them later, vMotion enabled or not, without a no-no dialog popping up.. (Not sure as Ive never done it as doing so ofc wouldn't make any sense...)

DRS is AFAIK a vCenter setting... not a VM or host as these are unaware of it as all DRS does is triggering vMotion ops when needed..

/Rubeck

[HemantPatel]

Hi,

vMotion is enable on Vmkernal Port. You can see the attached screen shot.

Manualy vmotion is working fine. But concern is only 2 host consuming more than 95% RAM still DRS is not moving VM automatically.

Interesting things is i set migration threshold to recommended by vmware on 3 and at that time it shows me load imbalance but if i change to 2 then it shows me Load Balanced but there is no change in automated migration by DRS.

In ESXi5.0 there is no need to create seperate vKernal port as there is no service console for management network so i have not created seperate vKernal port. Am I right here?

Do I need to create another vKernal port.

Regards,

Hemant 

[javierbardem]

Check Virtual Machine Options also, it is possible that DRS enabled in cluster level while disabled for particular VM.

Also check migration threshold, I am curious why current load deviation is not as high as expected.

[Rubeck]

You're right.. There's no need, but recommended IMO. Seperating the management from vMotion has always been best pratice..

You could try to add an extra vmkernel port (using a different subnet), and try to see if this helps although in theory it shouldn't make any difference...

Ohh, and try to flip the vCenter service first, althoug I bet you have tried this already...

/Rubeck

[HemantPatel]

Hi,

I checked VM option in under DRS setting and its is Default Full Automated level.

Regards,

Hemant

[HemantPatel]

Hi,

I restarted vCenter service to flush the cash stored in vcenter. Then after i simply refersh DRS and one of my VM configured with 12 GB RAM migrated to my other host. But Still one ESX is consuming more than 90% RAM and over allocated still DRS is not moving VM to my other ESX host which is having zero VM running.

Regards,

Hemant

[HemantPatel]

Hi,

I created new vKernal port as you sugested me with different subnet but using the same NIC which is used by management traffic. Let see what happen.

I am attaching screen shot for you.

Regards,

Hemant

[Slingsh0t]

If manually vmotioning a VM works fine, I wouldn't think that adding a second vmkernel port just for vmotion would make a difference.  DRS just instigates vmotions through vCenter if its on auto, it doesnt vmotion vms in any special way.

Can you turn your DRS threshold up to the most aggressive section and monitor for changes?

Have you got any affinity rules configured that may prevent DRS from shifting things to certain places?

[PUREJOY]

DRS algorithms are quite complex and it doesn't work as simple as say one ESX has max'ed out on its Memory and i need to move all the VMs to other host.

There are more criteria to automatically move VMs.

So what you are seeing is normal.

If you don't agree with that, you can take the "fully automatic mode" and set your own rules and apply them

_RV_

[HemantPatel]

Hi,

I have not configured any affinity rules which prevent DRS to make decision to move VM.

But as you told me i configured DRS threshold to most Aggressive and it start moving vm at the same time to my very least utilization esx server.

Now it balance the load on all my esx server.

Is this the only solution to set threshold to most aggressive?

Will this aggressive mode create problem if i will configure affinity rules in future?

Is this possible due to this aggressive mode there is too many migration in future?

regards,

Hemant