Our environment was Windows’s 6.0u3 app server, sql dB and external windows PSCs. We decided to make the jump to the appliance so we use the upgrade tool to also migrate us to the VCSA. I had already been using the web client full-time for two years, only using the thick client if something weird was happening with the web client. We also kept half of the environments on Windows 6.0u3g because of some comparability issues.
The web client started OK at first after the upgrade. Sometimes the flash plugin would stop responding (notice from Chrome) or the “<domain.com> not responding” message in IE. waiting it out and letting the plug-in load would usually be ok. I was actually able to install some SRM components so it wasn’t totally horrible. A few days later it did get to unusable.
We opened a case and tried a few things, clearing up SerenityDB, trying to delete a plugin directory. Restarting web services, disabling plugins via web client > administration > client plug-ins. Mostly this would give short term speed boosts but would ultimately go back to crappy performance.
At this point I started to think about memory usage. I looked at how much memory was on the VM and thought it seemed less then the old vCenter. Which is also weird because you would think the vCenter would be almost a combo of old vCenter plus old sql dB. The usage amounts in vCenter and in VAMI didn’t show an issue. I found this post in the forums where maxMB were more than assignedMB, so the person added more memory and the values of assignedMB increased (java heap memory assignments auto-increase as more memory is added to the VM and the vm is rebooted). The person didn’t specifically complain about performance, but it was a good place to start.
What I saw on my vCenter as that if I ran cloudvm-ram-size -S showed that maxMB was actually LARGER (almost 57GB) than the memory that was assigned on that VM (32gb). So my interpretation is that in effect Java actually needed more heap memory, and the maxMB was the demand of the memory for each process (and the total).
This is the output of cloudvm-ram-size -S on my VCSA:
TOTAL(RAM=32768MB) 12805 57207 57678 12606 16903 3307 8796093022207
So what I did was change memory to 64GB and once I rebooted the VM, the client was SUPER speedy. Back to 6.0 speed at least.
TLDR; monitor output of cloudvm-ram-size -S, is total maxMB is more then what is available on the VM, increase the VM’s memory.