ESPhome iPhones disconects all devices

Joe3 · December 17, 2021, 11:01pm

Looks like every time one of the four people in my house come home with there iPhone, all the ESPhome devices crash and become unavailable. Routers are all up to date. Anybody else have this problem?

tom_l · December 18, 2021, 12:13am

Do you have underscores in either your phone or ESP device names?

Are you using fixed IP addresses for your ESPs?

The only reason that I can think why this might possibly happen is if you have an mdns issue. Even then I’m not sure how this would occur.

github.com/esphome/feature-requests

Allow hyphens in node name

opened 10:57PM - 30 Sep 19 UTC

closed 10:02PM - 29 Jul 20 UTC

dannysauer

**Describe the problem you have/What new integration you would like** Allow hyphens in node name. **Please describe your use case for this integration and alternatives you've tried:** Node names currently allow underscore, which is in the set of valid mDNS service identifier and DHCP client id characters, but not in the set of (strictly) valid DNS hostname characters. Some people have DHCP servers which create dynamic DNS names, and DNS servers which are configured to strictly verify valid hostnames. Rather than altering network topology to accommodate mDNS, I would like to simply use DNS name resolution to reach ESP nodes. Cosmetically, I'd like to use a separator in the node hostnames. Hyphen is the only functional character for this. **Additional context** There are, realistically, other options for resolving this issue: * I could use a subdomain specific to the ESP devices rather than a naming convention. That's moderately inconvenient, as it means an additional configuration step rather than just letting the DHCP server hand out an address wherever on the network the device attaches * I could alter the DNS server to accept underscores. The drawback to this is that future devices could create hostnames with underscores, and underscores are not valid in real hostnames, dang it. :D * I could just mash the hostnames together instead of using hyphens. But that decreases legibility, and reduced legibility increases the possibility of mistakes later on. * I could rearrange my network so multicast DNS worked reliably. That's kind of a hassle, though. :) After a quick glance, it looks like this is just a matter of adding a hyphen to the string at https://github.com/esphome/esphome/blob/7ad593d674e488e3abbfb3208326e3cbcd404d24/esphome/config_validation.py#L46 - I'll wait to open a PR in a few hours, because Hacktoberfest. 👕

Either way it’s a weird issue and you will probably need some expert help. I suggest you post on the ESPHome Discord where some very clever folk hang out: https://discord.gg/P5rTy49w

mspinolo · December 18, 2021, 7:44am

Hi,

Had the same issue time ago and confirm it was due to mDNS storm.
My suggestion is to disable mDNS on all ESPHOME devices and from ESPHOME integration as first thing.

Then only way I know (but I am not an expert at all) is to use wireshark to understand where mDNS traffic is combing from and fix it.
In case you don’t want to dig in wireshark creating a VLan to segment your network (I would put ESPHOME devices on a single VLan and block mDNS on it) can help.

If your router support also dns proxy activation could help, but never tried as mine don’t.

Joe3 · December 18, 2021, 6:46pm

I’ll hook up wireshark and look for mass traffic during the shut downs. Now that the you guys mention monitoring I need to check out ntopng for major flow talkers during the event also. I’ll post my findings… Thanks guys.

Joe3 · December 20, 2021, 6:56am

Update… Even though when a iPhone entered the network everything failed, I did have random fails… So far everything looks cleaned up after spreading out 6 D1 esps that were all within 2 to 5 foot area in my office.