VMware Cloud Community
SyApps
Contributor
Contributor

Packet loss on all machines on network during vmotion

Hello,

I'm having a serious issue with packet loss across my entire network when I migrate a powered up VM from one host to another. I'm not just talking about losing pings within my VMware environment, I mean even physical computers outside of VMware will have "Request timed out" during the migration.

Vmotion is on the same network as everything else, we'll call that network vlan30. When I begin a migration, if I'm pinging my gateway from any computer on that network, I will get some packet loss for about 10 seconds. My vCenter server disconnects it's remote session, packet loss form vCenter to the gateway, packet loss from the machine I'm migrating to the gateway, even packet loss from some random physical server to the gateway.


What can I do to remedy this issue?

Always a big thanks to the community in advance! Dan Lee
0 Kudos
5 Replies
weinstein5
Immortal
Immortal

Do you mean all machines are sharing vlan 30? If so there are number of things wrong with this including that vmotion traffic is very bursty and is not encrypted that is why best practice to have vmotion traffic on its own network.

The remedy is to isolate the vmotion traffic.

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
0 Kudos
SyApps
Contributor
Contributor

Is there anything else as bursty as the vmotion network? Anything that would cause a few seconds of time outs here and there? We put vmotion on the same vlan for troubleshooting purposes. It will be off soon. What else would cause a few seconds of outages like that? Anything?

Always a big thanks to the community in advance! Dan Lee
0 Kudos
weinstein5
Immortal
Immortal

Not knowing what else if on your network there possibly could be something that is being bursty but since the problem starts when you vmotion and I assume stops when the vmotion is complete that the culprit is the vmotion traffic - 

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
0 Kudos
Madmax01
Expert
Expert

Hello theire,

So normally once DRS in use with Migrations threshold "middle"  > then VM's shouldn't balanced tooo much beetween the Hosts.  so because if you have too much activities > then this is a Indicator for having to less Ressources.  i have a Costumer with a Cluster.  it's running since 6-7 years and from the statistic theire where only 181 migrations for 61 vm's. so not much at all. (So its not a Mainplan to split vmotion). it's nice too have once having the Ressources > but not a Must have.

So during Migration the Switches for sure getting Pressure on the Network Layer.

- Which Esxi + vCenter version do you have?

- Do you see something on the Network layers regarding active Logs + Ressource Usage,...

- How big is the Memory from you're test VM?  same Problem with lower Memory Vm's ?

- do you have 1GB Full Duplex?

- How are you're vSwitches and Portgroups configured? Pnic's are all to same PSwitch like other Services?   ( So normally Virtualization Service should be splitted from other Dedicated Services,.. not good to mix it > could cause Problems and serious Conditions).

- Do you have WLB in Vlan30? Could cause also massive Problems.

So because strange is > if you tell that outside the Vmware Machines are loosing connections > then first it sounds for me that a Network Physical Part is having Problems to Calc the Load and has to less Power ,.... to much stressed,.... (First Idea without knowing the Environment)

and duplicate IP's i hope is not the Problem at all Smiley Wink.

Best regards

Max

0 Kudos
SyApps
Contributor
Contributor

- Which Esxi + vCenter version do you have?

4.0.0.x

- Do you see something on the Network layers regarding active Logs + Ressource Usage,...

Nothing logging, nothing in particular.

- How big is the Memory from you're test VM?  same Problem with lower Memory Vm's ?

Yeah, between 4-8GB each. Nothing serious there.

- do you have 1GB Full Duplex?

Yes

- How are you're vSwitches and Portgroups configured? Pnic's are all to same PSwitch like other Services?   ( So normally Virtualization Service should be splitted from other Dedicated Services,.. not good to mix it > could cause Problems and serious Conditions).

They are separate vswitches on each machine, with one vnice assigned to each vswitch, on

- Do you have WLB in Vlan30? Could cause also massive Problems.

Nope

So at this point, I'm just going to segment the vMotion network over to a different vlan and hope for the best. Thanks for your input.

Always a big thanks to the community in advance! Dan Lee
0 Kudos