Pull request 173: 5939-rm-healthcheck

Updates AdguardTeam/AdGuardHome#5939.

Squashed commit of the following:

commit fa01cf870db056313644a44ada844e4da61e8064
Author: Ainar Garipov <A.Garipov@AdGuard.COM>
Date:   Fri Jul 7 13:41:11 2023 +0300

    Docker: rm healthcheck
Ainar Garipov 2023-07-07 13:58:59 +03:00
parent e87acabfe5
commit b168fc4bba
1 changed files with 18 additions and 45 deletions

@ -157,18 +157,25 @@ program will overwrite them.
The settings are stored in the [YAML] format. The documentation describing all
configurable parameters and their values is available on [this page][conf].
NOTE: **Since v0.107.27** the container uses Docker-provided healthcheck
mechanism. The implementation uses special reserved domain name
`healthcheck.adguardhome.test.`, expecting it to resolve into NODATA answer. It
imposes restrictions on usage of this particular name, so specifying it within
the `blocked_hosts` array under the `dns` section of configuration file will
certainly break the healthcheck. For the same reason the `allowed_clients`
**should** contain and `disallowed_clients` **should not** contain the
`127.0.0.1` (`localhost`) address, which is used as a source of healthcheck
requests.
### `HEALTHCHECK`
[YAML]: https://yaml.org
[conf]: https://github.com/AdguardTeam/Adguardhome/wiki/Configuration
**Between v0.107.27 and v0.107.33,** the image used Docker-provided healthcheck
mechanism. It was causing many issues and has been removed **in v0.107.34.**
See issues [#5711], [#5713], and discussion [#5939].
If you need a healthcheck mechanism, it's better to create your own image
tailored for your configuration. Implementations may use the special domain
name `healthcheck.adguardhome.test.`, expecting it to resolve into NODATA
answer. It imposes restrictions on usage of this particular name, so specifying
it within the `blocked_hosts` array under the `dns` section of configuration
file will break the healthcheck. The `allowed_clients` and `disallowed_clients`
properties should allow the healthcheck client IP as well.
[#5711]: https://github.com/AdguardTeam/AdGuardHome/issues/5711
[#5713]: https://github.com/AdguardTeam/AdGuardHome/issues/5713
[#5939]: https://github.com/AdguardTeam/AdGuardHome/discussions/5939
[YAML]: https://yaml.org
[conf]: https://github.com/AdguardTeam/Adguardhome/wiki/Configuration
@ -227,37 +234,3 @@ your machine:
```sh
systemctl reload-or-restart systemd-resolved
```
## <a href="#known-issues" id="known-issues" name="known-issues">Known issues</a>
### Healthcheck
Since **v0.107.28** the container uses Docker-provided healthcheck mechanism.
If the implementation of the healthcheck script causes any issues with custom
Docker images and orchestration tools (like [#5711], [#5713]), then we recommend
disabling it by adding `--no-healthcheck` to the `docker run` command or using
your tool's equivalent.
(The actual change was made due to a necessity to handle zombie processes of
`wget` instances, see [PID 1 Docker problem][pid1]).
<!--
TODO(e.burkov): Fix the healthcheck for zeroes, update and uncomment this
section. Add the link to the issue.
### Listen addresses
The healthcheck script uses nslookup to check if the DNS server is up and
healthy. If the configuration contains `0.0.0.0` or `::` in the `bind_host`
property, the healthcheck will try to only check the `localhost.`.
Nevertheless, it will sometimes fail ([#5714]), the workaround is to configure
the actual IP address of the interface you want to listen on. Disabling the
healthcheck as described above should also work.
-->
[#5711]: https://github.com/AdguardTeam/AdGuardHome/issues/5711
[#5713]: https://github.com/AdguardTeam/AdGuardHome/issues/5713
[pid1]: https://blog.phusion.nl/2015/01/20/docker-and-the-pid-1-zombie-reaping-problem