I am aware Intel NUC isn't supported by VMware as a hardware to run on, but it does appear to be used a lot on home labs. I've come across an issue that appears to relate to e1000 network driver (according to Intel), and it involves the inability to wake the NUC machine via WOL (Wake-on-LAN) packet after it has been shutdown (via ESXi console or GUI). It works fine if you wake the machine up just after you switch the power on at the socket. But the minute you use a manual shutdown in ESXi, it will no longer wake up via WOL packet.
It has been raised here with Intel NUC:
But the conclusion by Intel is that it is caused by NE1000 driver VIB used in VMWare.
Anyone else have this issue, or have any ideas how I could potentially solve this?
If it's a problem with the ne1000 VIB that ships with 6.7 U1, you may try to get the ESXi650-201805217-UG patch and extract an older version of the same VIB from there and downgrade. Not sure that would work, but it's the next step to try (assuming you've already updated your BIOS to the latest published by Intel).
It's unsupported hardware to begin with, so I couldn't speak to any futures (especially because it is unsupported). It would seem a shame to return it because this one piece of functionality isn't present, but I suppose that's really up to you.
Indeed, it is a nice machine. I was asking because it appears a similar issue happened with a previous generation NUC but i could not find much, so i was wondering whether it was ever resolved...
Not sure. Also not sure about your use case, but consider that since they draw so little energy to begin with, one possible workaround is to just keep the thing running rather than shutting it down and attempting to wake it again.
In fact i forgot to mention the use-case. The nuc runs a number of VMs including my home firewall, so it is indeed always on.
The use-case is related to managing the UPS. The nuc shuts down when battery is low. However, if power returns before the UPS dies, the nuc (and other devices) might stay off. Therefore I have a small raspberry PI that checks. If power is available AND nuc is down, wakes it up. It has been working perfectly for years with an older nuc, and stopped when i upgraded the machine.
Sure it does work if the device is power cycled. The ESXi driver however puts the ethernet port in a state that does not respond to WOL anymore so when i shut it down nicely on UPS battery low, it is doomed. It is a recognised bug of the NE1000 driver, mentioned also on the intel web site.
What you may want to do is have your automation, upon low battery, perform a graceful shutdown of your VMs, and then just let it [the ESXi host] die to avoid that state. ESXi runs entirely from memory, and as long as the VMs are down gracefully there's really not a data consistency problem. It should be fairly trivial to add that.
i see your point but you dont see mine. If power returns before the UPS dies, nothing will restarts the VM and a manual action is required. As i run my home automation and many other things on that nuc, if this happens when i am not at home i lose all that.
Yes, I see your point, but I'm willing to bet there are workarounds. For example, setting a loop to check for UPS mains status, or eval on the mains status. Don't know what model of UPS you have but many of them have an ability to check for mains power. So it becomes and if/then situation. Just trying to think of other ways to help you out to bypass this WoL issue.
...and I thank you for taking the time! I am just saying that once the shutdown has started there is nothing I can do other than wake it up again, I guess (or am I missing something In your line of reasoning?)
It would really depend on how you're doing the shutdown. If the USB cable from the UPS is plugged into your NUC (ESXi host), then that obviously wouldn't work. But if it's plugged into your Pi (which is your automation station), then you could run all the logic there and call out to the ESXi host to gracefully shutdown all VMs. From there, two branches of logic would apply:
Either of these ensures that regardless of what happens after the UPS triggers an outage, your automation functions and keeps your VMs online. This is just how I would approach it myself in case it isn't what you're doing.