The problem A lot of people are having issues where their devices are not found. You either see a message like this:

ERROR: Could not find any Apple TV on current network

Or just empty results from atvremote scan. The issue can also come and go, so it works from time to time but very unreliably.

I do not know why this happens. I see it sometimes myself, but very seldom, so it’s hard to debug. So I am asking for help to solve this once and for all. If you have problems and want to help, continue reading.

Background Apple heavily uses Zero-configuration networking to publish and discover services on a network (they call it Bonjour). This simplifies user experience a lot, since no fiddling with IP-addresses or other network settings are required. Devices, like printers and media players, just “pop-up” when they are available.

The Apple TV works in the same way and pyatv tries to discover four distinct services:

• _touch-able._tcp.local. - Legacy protocol (DMAP) used by Apple TV 3 and earlier. Was supported until tvOS 13 was released as well. Requires additional pairing when used.
• _appletv-v2._tcp.local. - More or less the same as _touch-able._tcp.local. but is based on “Home Sharing”. If home sharing is enabled, the Apple TV can be controlled without pairing (simply put).
• _mediaremotetv._tcp.local. - This is the new “Media Remote Protocol” used by tvOS. Used by Apple TV 4 and later.
• _airplay._tcp.local. - AirPlay service used for instance when playing a video via AirPlay or performing screen mirroring. pyatv only use this for playing videos based on a URL.

Depending on which services pyatv finds, it will try to figure out the best protocol to use. If for instance both _touch-able._tcp.local. and _appletv-v2._tcp.local. are found, it will use the latter as no additional action is required by the user (i.e. no pairing).

If no service is found for some reason, then pyatv of course will conclude that no device exists. This is what happens for lots of users, including myself. pyatv uses a python library for zeroconf for service discovery and I suspect a bug in that library. By looking at open issues, other people report similar behavior as well.

What can I do? Troubleshooting this is tricky, especially since it’s not easily reproducible at all times. One thing I would like to conclude is if python-zeroconf is the culprit. We can do this with some confidence by scanning for Apple TVs with pyatv and also other tools on the same machine and network. If pyatv doesn’t find a device but avahi does, then we should suspect python-zeroconf.

So what you can do is:

1. Scan for devices with atvremote: atvremote scan
2. If the list is not empty, start over with 1) again a few times. If the device continues to be found, try again later to allow for caches to be flushed, etc.
3. When no device is found, try scanning with other tools. On Linux you can use avahi-browse like this:

$ avahi-browse -d local _mediaremotetv._tcp
+    br0 IPv6 Vardagsrum                                    _mediaremotetv._tcp  local
+    br0 IPv4 Vardagsrum                                    _mediaremotetv._tcp  local

or on macOS, you can do this:

$ dns-sd -B _mediaremotetv._tcp
Browsing for _mediaremotetv._tcp
DATE: ---Mon 14 Oct 2019---
10:50:35.015  ...STARTING...
Timestamp     A/R    Flags  if Domain               Service Type         Instance Name
10:50:35.016  Add        2   4 local.               _mediaremotetv._tcp. Vardagsrum

For Windows I don’t know. There probably are tools, but you will have to look that up yourself. 4. At this stage, if atvremote scan and the other tool you used did not find any device (and you have an Apple TV on your network), then there probably is a legitimate problem. You can report this, it’s useful information. 5. If pyatv did not find a device, but the other tool did, then we can start suspecting python-zeroconf. It would be good if you re-run the commands a few times to get some more entropy.

You can report any combinations you have tried. If you have more than one Apple TV and only one is found, that can be useful as well.

Note: The commands above only looks for MRP, so if you have an ATV3 or earlier and want to help out, change _mediaremotetv._tcp to _touch-able._tcp instead.

What can I do more? Provide additional logging from python-zeroconf during scanning. A simple script like this will scan for devices with debug logging enabled (it is not sufficient to use --debug with atvremote as that doesn’t enable debug logging for python-zeroconf):

import logging
import asyncio
import pyatv
import zeroconf

LOOP = asyncio.get_event_loop()


async def scan(loop):
    print(await pyatv.scan_for_apple_tvs(loop, timeout=5))


if __name__ == '__main__':
    logging.basicConfig(level=logging.DEBUG)
    LOOP.run_until_complete(scan(LOOP))

Please do note that the debug printout may contain some sensitive information. No passwords or such should be there, but things like unique identifiers and MAC-addresses might be present. Either you mask sensitive data out (which might be hard) or you just email me the dump (pierre.staahl AT gmail.com). But remember, it is up to you if you want to share the log. I will only use it for debugging.

Great, so anything else? There is another (asyncio) zeroconf library for python that might be interesting to try out. It’s called aiozeroconf and is available here:

https://github.com/frawau/aiozeroconf

You could try installing that and scanning for services as well, maybe it works better? It’s still based on the same code base as python-zeroconf, so no guarantees.

Instructions:

aiozeroconf -s _mediaremotetv._tcp.local.

Browsing services, press Ctrl-C to exit...

Service Vardagsrum._mediaremotetv._tcp.local. of type _mediaremotetv._tcp.local. state changed: ServiceStateChange.Added
  IPv4 Address: XXX
  IPv6 Address: YYY
  Weight: 0, priority: 0
  Server: Vardagsrum-2.local.
  Properties are:
    ModelName: Apple TV
    AllowPairing: YES

Can I do EVEN more? If you are an experienced user, you can do some packet interception and look the broadcasted data. I have taken dumps with tcpdump (and also looked at zeroconf debug data) when no devices were found by pyatv and made the conclusion that no response was received at all (on network layer) when this happened. It basically means that no response was received by python-zeroconf, so it wasn’t just ignored for some unknown reason. It would be interesting to tap in on network traffic from another computer (with wireshark or so) and see if the expected data it received there or not. It would also be interesting to compare what python-zeroconf does, network-wise, and compare to other tools (like avahi and dns-sd, mentioned earlier). Maybe something is not configured correctly?

I leave this as an open bullet for experienced used to fiddle with.