Hello,

vmware-vpxa is installed only when connected to VC. To date I never had any problems when restarting this service. VMs stay alive and working because VMs do not depend directly from VC but from the ESX Server.

I think you can even stop this service without causing any damages on VMs.

Thank you

Dominic

Had this problem with a older esx (2.5.2) host, in that case the server was running out of diskspace.. can check this with the 'df -h' or 'vdf -h' command

service mgmt-vmware restart
service vmware-vpxa restart

Will not kill your running VMs. If you are trying to Re-sync with the VC, stop the VC service (on the server running VirtualCenter), restart these 2 services and then start the VC service. We have found that this will normally re-sync the ESX host with the VC.

the only service restart that MIGHT cause a VM problem is mgmt-vmware if you have an older release of 3.5 or before, or you have Automatic Startup/Shutdown enabled on the host.

verify that you can use the VI (or vsphere) client to connect directly to the host (use host IP/name in address field and root for username). this goes directly to the host so vpxa's status doesn't matter. If you can login this way, you can manage your vm's (but no vmotion as that's a vc function) and then you can narrow down the possibible problem causes. If you can't get in this way, then hostd(mgmt-vmware) isn't staying running, the watchdog is confused or there is an issue preventing hostd from operating properly.

You can check the first two possibilities by running 'ps -ef|grep hostd', if you only get back the process for grep hostd then you probably need to delete the watchdog PID file. If it is running, check /var/log/vmkwarning and vmkernel for errors. Usually it's storage related problems that cause this.

HTH