Fork me on GitHub

Background: Wake on Lan support was asked for sometime ago and it almost delayed Webconverger 13.0 until it was moved to target 14.0 and it's one of those trivial new features that has cost at least 3 days worth of man hours of development time.

Why isn't wake-on-lan, simply not just the default?

I think it is the default on some machines but not others. Webconverger shipping ethtool seems like just a waste of space, to configure silly machines to flip that bit.

For example after installing ethtool for the first time on a Lenovo Thinkpad X220 and I get on a sudo ethtool eth0 invocation:

Supports Wake-on: pumbg
Wake-on: g

So we are fine. So by default my X201 was "Wake-on-Lan" ready, if there is such terminology.

BIOS override

I noticed on a machine that refused to wake up, that it was in fact disabled in the BIOS. DOH. Is this reflected on the Linux user space tool ethtool? NO. Very annoying.

Shutdown now not working

I noticed that once enabling Wake-On-LAN in the BIOS, a certain Intel based machine refuses to cleanly shutdown for whatever reason. :(

When it does work, it can do strange things

I managed to Wake-On-LAN my X201 and then it tried to netboot for about a minute before booting. Why ....


Which port ?!?

Oh btw, there seems to be a million ways Wake-on-LAN is implemented/triggered. I've chosen the sanest Magic Packet flavour which needs a mac address in order to work.

Is it port 40000 or port 9? I can't find a definitive answer for Magic Packet since the normal wol utility goes on port 40000 by default for whatever reason.

Update: Port doesn't matter, the hardware just looks for any packet with the magic sequence

And security?

Tbh in Enterprise network level security is usually well done, so I really don't see the need for passwords to be set. Especially the way most client connections are behind a NAT nowadays. Daft.

Horrible usability of wol

Say you are testing from your laptop, sharing internet on eth0 from You need to do an invocation like:

wol -v -p 9 00:04:5f:04:bf:dd -h

Not many people know "255" is a broadcast address. Sigh

Is Wake-on-LAN the right way to manage machines power usage?

In "Enterprise" environments having timed shutdowns and ways to wake the machines in the morning seems like the "old" approach for Energy saving. Timed shutdowns for example are obviously NOT very desirable if you actually working on a Webconverger machine after hours.

For example Apple Mac machines aggressively save power at all times unless for example you are using the machines or prefix commands with caffeinate.

This "sleep always until actually used" approach seems better for real energy savings for kiosks. Nonetheless lots of Webconverger customers want noblank option so that their kiosk users see and know the machines are ready for use. Just imagine walking up to a blank screened library kiosk. Would you know that pressing the keyboard would wake it up? Not everyone would know this without a sign...

For Web signage, a different use case, a different approach needs to be taken again to keep the sign visible at all times. In our deployment we dim the screens from midnight to morning and the machine is never halted and currently this is controlled on a hardware level.