The below was the customer environment:
- Two PVS 5.6 SP3 Servers on HP BL460c G7
- Windows Server 2008 R2
- Eight Target Devices (HP BL460c G7)
- Streaming Windows Server 2008 R2
- XenApp 6.5
- Windows License Server Version 11.10.0 build 12012
When the customer manually restarted the hanging machines they would restart fine 95% of the time.
The customer provided me some nice wireshark screenshots showing the point at which the boot failed but as he had filtered on the IP.Addr I suspected we had missed something.
We re-took the trace and discovered that by filtering on the mac address instead we could see the following:
Note how the PVS server suddenly loses its connection and ARP entry for the target.
Suspecting something low level we had a look at the NIC firmware to find it was version 4.0.493.0.
Looking at http://support.citrix.com/article/CTX128834 we would think that the NIC firmware of this version should be OK but to be sure we decided up upgrade to Firmware 4.1.450.7
On the new version the issue was gone and after a week of testing the case was another in the closed pile.
The lesson: for PVS wireshark is crucial.... as are the filters you use on the traffic it captures.