Recently, I was working on an UCS blade firmware upgrade along with esxi upgrade from esxi 5.5 to 6.0 and came across this error where the esxi host became unresponsive with an error “can’t fork” on its DCUI.
here is a little background on this story, this particular blade was B240 blade which was being used as SAP HANA blade by the customer and the firmware upgrade and esxi upgrade went fine and two days later the host became unresponsive and we couldn’t connect to it using SSH, DCUI, etc, connecting to the kvm console revealed the below screen when we went to its Alt+F1 command interface
we had to bounce the box and we had to reduce the linux vm memory which was hosting SAP HANA on it to be 10% less than the memory of the esxi host.
Conclusion: The HANA VM (linux) on the esxi host should have 10% less memory than the overall memory of the esxi host to avoid this problem.
Converged, Hyperconverged Infrastructure & Virtualization 