Ran into a fun issue after updating our VMWare environment to 7.0.3.00500. vCenter updated cleanly, and so did the hosts, but after an hour my hosts started to disconnect from vCenter and would not re-connect. Did all the normal troubleshooting for a few hours before getting VMWare support involved.
They saw multiple memory leaks happening on the hosts and initially suspected a bad patch. Digging deep into the logs though we discovered an interesting "bug". On the hosts the hostd service is used to connect to vSphere and in the vSphere host that request in handled in heap memory of vsphere-ui service. The default value for this is 1024, which is not enough. The fix is to double the heap memory size and then you can re-connect the hosts. (Or increase this value prior to updating)
source for increasing heap memory: Manually increasing the heap memory on vCenter Server components in vCenter 6.x / 7.x (2150757) (vmware.com)
Use SSH to connect to the VCSA and enter the Bash shell
Run the following command to output memory assigned to each of the services:
cloudvm-ram-size -l
Increase the amount of memory assigned to the services by running the following command, using the vCenter HTML5 web client (or "vsphere-ui"), as an example:
cloudvm-ram-size -C XXXX vsphere-ui
Note:
XXXX should be replaced by the desired amount of memory in MB.
For example, cloudvm-ram-size -C 2048 vsphere-ui would increase the memory of the vsphere-ui service from 768 MB to 2048 MB.
Where there is no one size fits all memory value to set in this scenario, as a guide, double the size of currently assigned memory of the service that has an issue. This may need to be repeated.
Restart the service that has been altered, in this example vsphere-ui:
service-control --stop vsphere-ui;service-control --start vsphere-ui;
Confirm that the service has now been allocated the increased memory using the command:
cloudvm-ram-size -l