[Bug] Filewatcher stops watching dns.extra_records_path

[nblock]

Is this a support request?

• This is not a support request

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

A read-edit-write cycle (with vim) causes headscale to lose track of the path in dns.extra_records_path.

Expected Behavior

Pickup changes made to the file. Headscale logs detected writes with log.level: trace when different versions of the config file are written. The sequence:

cp extra1.json .headscale-dns/extra.json
cp extra2.json .headscale-dns/extra.json
cp extra2.json .headscale-dns/extra.json    # no changes, OK

yields (fine)

2024-12-14T16:39:47+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=WRITE path=.headscale-dns/extra.json
2024-12-14T16:39:47+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:134 > extra records updated from path, count old: 1, new: 1 records=[{"Name":"grafana.myvpn.example.com","Type":"A","Value":"100.64.0.1"}]
2024-12-14T16:39:56+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=WRITE path=.headscale-dns/extra.json
2024-12-14T16:39:56+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:134 > extra records updated from path, count old: 1, new: 1 records=[{"Name":"grafana.myvpn.example.com","Type":"A","Value":"100.64.0.2"}]
2024-12-14T16:40:01+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=WRITE path=.headscale-dns/extra.json

Steps To Reproduce

The following causes headscale to loose track of the file:

vim .headscale-dns/extra.json    # make some valid changes and save

yields sometimes:

2024-12-14T16:50:21+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=RENAME path=.headscale-dns/extra.json
2024-12-14T16:50:21+01:00 ERR ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:115 > reading extra records from path: .headscale-dns/extra.json error="reading path: .headscale-dns/extra.json, err: open .headscale-dns/extra.json: no such file or directory"

and yield sometimes:

2024-12-14T16:51:45+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=RENAME path=.headscale-dns/extra.json
2024-12-14T16:51:45+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:134 > extra records updated from path, count old: 1, new: 1 records=[{"Name":"grafana.myvpn.example.com","Type":"A","Value":"100.64.0.7"}]

Result is always the same:

• Headscale no longer sends DNS updates to nodes - a query yields stale results
• Changes to the file are no longer picked up. Sometimes an error message is emitted and sometimes not
• Headscale needs to be restarted for the filewatcher to work again

---

This behavior can be triggered reliably with:

mv extra.json .headscale-dns/extra.json 

which yields:

2024-12-14T17:47:51+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=CHMOD path=.headscale-dns/extra.json
2024-12-14T17:47:51+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:134 > extra records updated from path, count old: 1, new: 1 records=[{"Name":"grafana.myvpn.example.com","Type":"A","Value":"100.64.0.1"}]
2024-12-14T17:47:51+01:00 TRC ../runner/work/headscale/headscale/hscontrol/dns/extrarecords.go:87 > extra records received filewatch event op=REMOVE path=.headscale-dns/extra.json

After that, changes to the file are no longer picked up.

Environment

- OS: Arch
- Headscale version: 0.24.0-beta.1
- Tailscale version: -

Runtime environment

• Headscale is behind a (reverse) proxy
• Headscale runs in a container

Anything else?

No response