Every few minutes "Host is not responding"

[biokovo]

Hi.

Two esx5i U1 servers, Linux vCenter (vCSA), esxi is installed on flash card, shared SAS storage.

Every 11 minutes I get a message "Host is not responding" on booth servers, and after a few seconds everything is ok. There is no impact on VM's.

I tried many things and don't have no more idea.

Thanks

[nielse]

Have you considered checking the logs? If so are there any errors in the logs (vpxa.log, hostd.log,...)?

[biokovo]

Thanks.

I check the logs, but I can not find out the reason.

In hostd.logs there is every few minutes (not at the time when hosts disconected) errors:

[FFF81B90 verbose 'Statssvc'] HostCtl Exception in stats collection: Sysinfo error on operation returned status : Not found. Please see the VMkernel log for detailed error information

vCenter and hosts are not in the same subnet but I don't think it is network problem.

I check out DNS names, time, and really don't know where to look.

[nielse]

It could be the network problem tho. Did you check the vmkernel.log for errors?

Can you try putting 1 host in the same network and add it to your vCenter? Does this host also disconnect?

[Sreejesh_D]

check the Tasks& events of the hosts and see what was going on before host disconnected.

We've faced host disconnect issue recently due to excessive DRS initiated vmotions.

[biokovo]

I can't see anything in task&events related with this.

I don't think DRS/vmotions or something similar couse this. Like I said, this happens every 11 minutes without any evident reason.

In vmkernel.log there is no error.

A few days ago I update hosts and restart them but the problem still exist.

[Sreejesh_D]

check vpxd.log at vCenter server for any errors or exception. Usually when a host disconnects there will be an entry in vpxd.log.

[vGuy]

Are your ESX hosts behind NAT. If so, you will need to add the NAT'd addr. to vpxa.cfg file on each host.

Also, please verify all the ports required for vCenter -> host communication are allowed bi-directionally. A good test would be to allow all the ports temporarily and test..