OctoPi 0.13 does not work with Pi Zero W

[mitsumaui]

When installing OctoPi 0.13 to SD card, no Wifi adapter is detected on the Pi Zero W. This is built in now, so expect an updated Raspbian image is required? Will test further with Raspbian.

[guysoft]

1. It will not work because the Pi Zero W was released after 0.13. 2. The raspberrypi themselves didn't get it working yet. If you look at the release notes ( http://downloads.raspberrypi.org/raspbian/release_notes.txt ) you will see it says "initial support for Pi Zero W". Aka its still initial and we are expcting them to release a version with fixes. I will get the nightly coming out hopefully tomorrow, the wifi in my house has been acting up and breaking the build server.

[mitsumaui]

Thanks for the response! I saw this in the Raspbian rel notes so assume it's mostly ready to go:

2017-03-02:

• Updated kernel and firmware (final Pi Zero W support)

I've just installed octoprint on the latest Raspbian Lite image manually to see how it goes! Once you have this into nightlies - let me know (along with a brief instruction on how I can load to SD card for testing) and I can do some testing.

[guysoft]

Huh, they changed it. It said "initial" a few days ago.
Oh well, @foosel already got errors when trying the camera. And we are on the move for 0.14.0 RC, when we are ready.

[foosel]

Some more infos on the errors I encountered...

Utilizing actual bandwidth on the WiFi interface (as in, streaming from the webcam in the "Control" tab in OctoPrint and also directly via /webcam/ plus doing a simple speed test for downloading something via wget) I saw huge CPU consumption by kworker and irq in the output of top. And printing a tri spinner (lots of curves) and trying to watch the webcam resulted in endless stuttering thanks to the above mentioned processes eating up ~50% of all available CPU resources and the constant swapping between processes not leaving enough air to breathe for OctoPrint to feed data fast enough to the printer.

That was still against the "initial support" image, aka 2017-02-16 though. I did a manual rpi-update on that which pushed the kernel version to 4.9 and reduced the CPU load during WiFi usage to something around 25%. Still some stuttering but at least better.

I then installed the SnapStream plugin and used that configured for 4fps instead of the actual webcam stream. That brought the CPU usage down to 15% and the print went through.

Still... a decidedly "meh" experience so far. As a side note, I also tested against a Pi3 then out of curiousity: same behaviour but doesn't matter thanks to four cores being available.

tldr: Actually using the WiFi of the RPi Zero W to transfer data (= watching a webcam stream of a print) eats a lot of CPU and can severly impact achievable print quality.

[foosel]

Some quick shots:

This is only one browser connected to http://octopi.local/webcam/stream.html. OctoPrint server isn't even running.

And this is a simple wget -O /dev/null http://somehost:someport/somelargefile.gcode executed on the Pi.

And look at that load average. Hilarious. As a reminder, we are not doing anything computational intensive here. Just pumping some data through the wifi interface.

[nophead]

I see similar things with RPI B. I don't know how anybody successfully uses Wifi and a camera with only one core. I have to use Ethernet.

…

On 7 March 2017 at 16:23, Gina Häußge ***@***.***> wrote: Some quick shots: [image: image] <https://cloud.githubusercontent.com/assets/83657/23665870/fecaa7a8-0359-11e7-8760-05b50aad71ec.png> This is only one browser connected to http://octopi.local/webcam/ stream.html. OctoPrint server isn't even running. [image: image] <https://cloud.githubusercontent.com/assets/83657/23666103/926ea81a-035a-11e7-9b7b-6afb7a43a8ed.png> And this is a simple wget -O /dev/null http://somehost:someport/ somelargefile.gcode executed on the Pi. And look at that load average. Hilarious. As a reminder, we are not doing anything computational intensive here. Just pumping some data through the wifi interface. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#318 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAijhcNxSkdtSkdFN1yAl6nED0cUmYMWks5rjYSSgaJpZM4MTFtP> .

[foosel]

@nophead I think I already mentioned that once, but I never had such issues with USB wifi dongles. This mmc stuff though...

Now testing one of my dongles with the RPi Zero W.

[foosel]

@nophead now I finally see what you mean with CPU load by haproxy ;)

That's a running print with the Control tab (and hence the webcam stream) open. USB dongle, internal NIC is disabled. Slight stuttering but only at the start. Couldn't let it run (appointment) so no idea how that would develop over time. Next to no lag though (same setup with internal nic had lots of lag).

That is only streaming from /webcam/, OctoPrint server is disabled. USB nic, internal one disabled.

And wget.

Needs more tests but current gut feeling based on observed print behaviour: USB WiFi dongle performs better than built in interface.

[guysoft]

Next nightly would be with the latest base image

[pposthoorn]

Newest build works for me with the new raspberry Pi zero W with Pi camera.

[calonmerc]

Any additional testing needed on this? I've got a Pi Zero W doing nothing (but no Pi cam, still in the mail).

[foosel]

More testing is always good!

@guysoft i released OctoPrint 1.3.2 yesterday, so that is now ready for inclusion in 0.14 as well (and might have already made it into this night's nightly? Not sure if that is build against master, would make sense though)

[guysoft]

nightly build is against master. If you got it in before midnight UTC it should be there

[foosel]

Ah, great. I wasn't sure anymore if the nightlies were master or devel branch. Release was 3pm CET/2pm UTC so that should have made it in :)