Network connection stops working unexpectedly for one or several Windows virtual machines on the node.
The following message can be seen in the "
/var/parallels/$VM_NAME.pvm/parallels.log" around the same time network connection is lost:
1640355.492 sec. ERROR: e1000_rx_write: prev_size > new_size (255 > 0). rdh 148, prev_rdt 147, new_rdt 148, size 256 l_rdh 148 l_rdt 147
The problem is caused by a race in e1000 driver inside the VM. It could happen under excessive loads on the server.
The problem will be fixed in the scope of the internal issue with ID PSBM-26250. The fix will be included in on of the future product updates.
As a workaround, switch the virtual machine's network adapter to Realtek:
# prlctl set VM_NAME --adapter-type rtl
The change requires a reboot of the virtual machine.