Clients not found by Omada

[stordoff]

This seems to be an Omada bug, but I'm wondering if there's a way to fix it/a workaround with coredns_omada. Fairly frequently, wired hosts disappear from the Omada client list (seemingly those with not much traffic). When this happens, coredns_omada can't resolve the client:

Expected:

>nslookup alpine-docker.int.stordoff.com 10.0.0.5
Server:  alpine-docker.int.stordoff.com
Address:  10.0.0.5

Name:    alpine-docker.int.stordoff.com
Address:  10.0.0.5

Actual:

>nslookup alpine-docker.int.stordoff.com 10.0.0.5
Server:  UnKnown
Address:  10.0.0.5

*** UnKnown can't find alpine-docker.int.stordoff.com: Server failed

alpine-docker is the host when the coredns_omada Docker image is running, and it has IP address 10.0.0.5 (static DHCP lease). Once the client reappears in the client list, it can be resolved again.

[dougbw]

Hey, this was something that crossed my mind as currently every time the refresh is run a new set of records is created, which can cause records to be removed quite aggressively (I also have a particular host which seems to flap in and out of the Omada controller).

It would be a straightforward change to start the refresh from the existing set of records, but I would need to have a think about how to then cleanup stale records after a period of time, as keeping the records indefinitely could cause some incosistenties.

[stordoff]

Hi,

Thanks for replying. I don't know if it's possible or would cause other issues, but just keeping the DHCP reservations available indefinitely would resolve most of the issue for me. Any other hosts disappearing/reappearing are not a major concern (I assume I could add these reservations manually to coredns, but that defeats the purpose of keeping management within Omada).

[dougbw]

I could put the option to keep stale records behind an optional config property which would be quite a simple change. I will take a look the next time I am working on this.

[dougbw]

I have had a look at the code and am planning to make changes to support a couple of features.

First of all I will add resolution for DHCP reservations as managed by the Omada controller in the services page.

Secondly rather than starting from a fresh list of clients every refresh it will retain state across refresh intervals, whilst purging records once they exceed a configurable age (e.g 1 hour). This will make the refresh slightly less aggressive as I have had minor issues with records disappearing briefly if a client crashes/reboots/etc.

[dougbw]

I have just published a beta release with support for resolving DHCP reservations and persisting stale client/device records across refreshes. By default stale records are purged after 5 minutes (although this can be extended via config).

• https://github.com/dougbw/coredns_omada/releases/tag/v1.7.0-beta
• https://github.com/dougbw/coredns_omada/pkgs/container/coredns_omada/296621837?tag=1.7.0-beta

I am going to do some more testing before promoting this release, but initially it looks to be working fine.

[stordoff]

Thanks for that. I've pulled the beta Docker image (trying stale_record_duration 60m as a few of my clients disappear for extended periods) and it seems to be working fine for now. I'll let you know if I run into any issues.