Vmotion taking a long time

[Mllii]

Some of my VM are taking a long time to migrate to another host in the cluster over 2+ hours.

this is holding up other tasks as well.

the the host are connected by SAN and this happens with both DRS mirgrations and ones i haves set to migrate my self.

[Fabio_Pitzolu]

Hi :smileyhappy:

The Vmotion process is based on moving resources between 2 ESX hosts, memory and cpu.

The big job is moving all the memory pages over the network between the hosts. It can be slow for some reasons:

\- The amount of memory to migrate;

\- How much memory is in used by the VM while migrating;

\- The use of the swap file;

\- The speed of the network link.

VMotion reccomendation for network is to have a Gigabit link, do you have it?

You should also check how much memory is used by your VM while migrating, also the swap file usage.

Post these details so we can have a deep look and try to help more.

Hope my terrible english doesn't bothered you :smileyhappy:

Bye!

Fabio

[Mllii]

\- The amount of memory to migrate; The guest OS memory is 2gig

\- How much memory is in used by the VM while migrating; guest usage 50mb

\- The use of the swap file;

how would i find out that?

\- The speed of the network link.

all the networks are at 1gb including Vmotion.

this is happening to random VM in my cluster of 14 servers?

[MayurPatel]

Under normal conditions it should not take 2 hours to migrate a VM using VMotion.

I would check the VMotion networking setup, it is possible you may have a problem with NIC negotiation on the physical switch and Host NIC ie Full/Half Duplex mismatch.

[hihiy]

VMotion action is base on the resources of the ESX Hosts. And also, please check your network problem.

[Michelle_Laveri]

Could it be a network speed/duplex problem?

OR a faulty NIC?

Regards

Mike

[hihiy]

One more question: Your VMotion takes long time. Is VMotion task sucessful or failed???

I think the key is the ESX server resouces.

[Mllii]

the task timed out after 4 hours and 30mins.

[Texiwill]

Hello,

Are all 14 servers connected to the same LUNs? It is very important to verify your network connectivity and we we need to look at some logs as well.

A few things:

1) is your vMotion network a private physical switch/VLAN only used by ESX servers, or is this on a public network.

2) Are you routing the vMotion between two networks (routed)?

3) Please post output of:

esxcfg-vmknics -l

esxcfg-route -l

esxcfg-nics -l

vmkping vMotionIPofTargetHost

4) review /var/log/vmkernel for any errors regarding the vMotion.

Best regards,

Edward

Message was edited by:

Texiwill

[Mllii]

Checking all the vMotion nics on the servers on of them was flicking between 100 and gig an that was what was causing to issue.

i got the network guys to look at it and hopefully we will work out why.

im guessing faulty port

[GDCM]

Sorry to just hop in, had a quick thought...how do you have the vnic for the vmkernel (vmotion) configured? is it auto negotiate or is it set to 1GB Full? I ask because I've rarely seen a port problem where it flips between 100 and 1000 but I've seen tons of instances where the auto negotiate doesn't work to well and causes flipping. Just a thought.

[bertdb]

on the service consoles of the ESXes involved, can you vmkping the vmotion addresses of every other ESX server ?

If one of these vmkpings fails, there is a networking issue between the different VMkernels. That would certainly break VMotion.

(as Texiwill suggested above, I see now.)