Dockerfile does not have health check #3290

agneevX · 2021-06-28T07:45:40Z

Continuation of #1426

There's no health check for Docker images.

This is how Pi-hole does it:

HEALTHCHECK CMD dig +norecurse +retry=0 @127.0.0.1 pi.hole || exit 1

agneevX · 2021-06-28T07:49:35Z

Using curl for web UI:

HEALTHCHECK CMD curl -fs http://localhost:3000 -o /dev/null || exit 1

ainar-g · 2021-06-28T12:52:21Z

Unfortunately, it's not that straightforward. AGH's initial HTTP API port is 3000, but the user can change it. In fact, the port that we currently propose by default is 80. And people can change it in the config file. We could include some sort of a script that would read the config file and use that value, but definitely not in this release cycle.

sachinwadhwa · 2021-06-28T13:50:10Z

I have used following healthcheck in my compose file. Port 53 is most important for me (as I am not using DoH).

healthcheck:
  test: "/bin/netstat -pant | /bin/grep 53"
  interval: 45s
  timeout: 30s
  retries: 3

ameshkov · 2021-07-05T11:33:06Z

DNS server is not working until the initial setup is finished

agneevX · 2021-07-12T15:36:10Z

Seems like a separate script is needed to accomplish this.

lolgast1987 · 2021-08-03T18:14:50Z

I made something to do exactly that. Checks for the Setup to be done and if so, which port is configured for bind_port.

https://github.com/lolgast1987/adguard-unbound/blob/master/files/healthcheck.sh

PeterDaveHello · 2022-09-07T16:32:26Z

Can we at least accept to use maybe 53 port with nslookup or dig to do a basic health check? I'd like to send a pull request for it if it's acceptable. We could have some further improvement later, when the more complete and complex idea came up?

PeterDaveHello · 2022-09-07T16:53:26Z

My current healthcheck config using docker-compose:

https://github.com/PeterDaveHello/dnslow.me/blob/master/docker-compose.yml#L9-L14

healthcheck:
  test: nslookup www.google.com || exit 1
  timeout: 5s
  interval: 60s
  start_period: 10s
  retries: 1

JaneJeon · 2023-02-13T20:28:10Z

For those of us using DNS-over-HTTPS, what is the best option here?

PeterDaveHello · 2023-02-14T08:57:42Z

For those of us using DNS-over-HTTPS, what is the best option here?

https://github.com/PeterDaveHello/dnslow.me-uptime/blob/master/.upptimerc.yml#L8

You can use http get method to perform a query like https://dnslow.me/dns-query?dns=AAABAAABAAAAAAABBmRuc2xvdwJtZQAAAQABAAApAgAAAAAAAFoADABWAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

JaneJeon · 2023-02-14T19:02:43Z

What in the world is the random string of letters in the ?dns= query and how do I construct it myself?

PeterDaveHello · 2023-02-14T19:11:18Z

What in the world is the random string of letters in the ?dns= query and how do I construct it myself?

It's not random, I just do it the lazy way, use any DoH client to send a http get query to a https server, then you can find it in the access log.

zoonderkins · 2023-02-23T07:18:38Z

Another example to perform www.example.com

https://doh-jp.blahdns.com/dns-query?dns=AAABAAABAAAAAAAAA3d3dwdleGFtcGxlA2NvbQAAAQAB

alexsannikov · 2023-03-09T20:12:41Z

As an example: Node Red docker container uses healthcheck....

Merge in DNS/adguard-home from 3290-docker-healthcheck to master Updates #3290. Squashed commit of the following: commit 3ac8f26 Merge: bc17565 0df3260 Author: Eugene Burkov <[email protected]> Date: Mon Mar 27 01:09:03 2023 +0500 Merge branch 'master' into 3290-docker-healthcheck commit bc17565 Author: Eugene Burkov <[email protected]> Date: Sun Mar 26 18:04:08 2023 +0500 all: fix script commit e150fee Author: Eugene Burkov <[email protected]> Date: Sun Mar 26 17:18:12 2023 +0500 all: imp naming commit 26b6448 Author: Eugene Burkov <[email protected]> Date: Sun Mar 26 03:13:47 2023 +0500 all: support https web commit b5c09ce Author: Eugene Burkov <[email protected]> Date: Sat Mar 25 20:03:45 2023 +0500 all: imp scripts fmt, naming commit 8c3798c Merge: e33b0c5 fb7b8bb Author: Eugene Burkov <[email protected]> Date: Sat Mar 25 00:25:38 2023 +0500 Merge branch 'master' into 3290-docker-healthcheck commit e33b0c5 Author: Eugene Burkov <[email protected]> Date: Fri Mar 24 16:47:26 2023 +0500 all: fix docs commit 57bfd89 Author: Eugene Burkov <[email protected]> Date: Fri Mar 24 16:44:40 2023 +0500 dnsforward: add special-use domain handling commit f04ae13 Author: Eugene Burkov <[email protected]> Date: Fri Mar 24 16:05:10 2023 +0500 all: imp code commit 32f150f Author: Eugene Burkov <[email protected]> Date: Fri Mar 24 04:19:10 2023 +0500 all: mv Dockerfile, log changes commit a094a44 Author: Eugene Burkov <[email protected]> Date: Fri Mar 24 04:04:27 2023 +0500 all: finish scripts, imp names commit 4db0d0e Author: Eugene Burkov <[email protected]> Date: Thu Mar 23 18:33:47 2023 +0500 docker: add script and awk program

Merge in DNS/adguard-home from 5642-fix-healthcheck-ssl to master Updates #5642. Updates #3290. Squashed commit of the following: commit c457ecb Author: Eugene Burkov <[email protected]> Date: Mon Mar 27 15:35:32 2023 +0500 docker: imp docs commit fddabb9 Author: Eugene Burkov <[email protected]> Date: Mon Mar 27 15:18:22 2023 +0500 docker: skip ssl check

EugeneOne1 · 2023-03-27T11:34:32Z

Hello again, @agneevX, and everyone here. The latest edge build now implements the healthcheck via Dockerfile. Could you please check if it works for you?

stavros-k · 2023-03-27T16:28:37Z

@EugeneOne1
Hello, healthcheck does not account for the --port flag when it's not yet run through the wizard.
https://github.com/AdguardTeam/AdGuardHome/blob/132ec556dc206c7f4453b194d58112944b5b96db/docker/healthcheck.sh#LL25C27-L25C31

EugeneOne1 · 2023-03-28T11:37:37Z

@stavros-k, thanks for the feedback. Actually, the current implementation expects the port 3000, specified in the official Dockerfile, to be used for installation. Since this use case exists, we've filled the separate issue (#5645) about taking custom ports specified via command-line options into account.

Are there any other issues happened?

stavros-k · 2023-03-28T12:30:33Z

@EugeneOne1 I haven't noticed anything else yet. But currently working on adding it as an app to TrueNAS SCALE. I'll let you know if I find anything

leo15dev · 2023-03-29T13:48:24Z

@stavros-k, thanks for the feedback. Actually, the current implementation expects the port 3000, specified in the official Dockerfile, to be used for installation. Since this use case exists, we've filled the separate issue (#5645) about taking custom ports specified via command-line options into account.

Are there any other issues happened?

If the health check status showing unhealthy, it will produce a lot of zombie process.

e.g.
root 119494 0.0 0.0 root 119495 0.0 0.0 root 119939 0.0 0.0 root 119956 0.0 0.0 root 120419 0.0 0.0 root 120420 0.0 0.0 root 120900 0.0 0.0 root 120901 0.0 0.0 root 121332 0.0 0.0 root 121341 0.0 0.0 root 121816 0.0 0.0 root 121817 0.0 0.0 root 122284 0.0 0.0 root 122285 0.0 0.0 root 122792 0.0 0.0 root 122793 0.0 0.0 root 123281 0.0 0.0 root 123282 0.0 0.0 root 123721 0.0 0.0 root 123722 0.0 0.0 root 124209 0.0 0.0 0 0 ? Z 19:01 0:00 [ssl_client]
0 0 ? Z 19:01 0:00 [ssl_client]
0 0 ? Z 19:01 0:00 [ssl_client]
0 0 ? Z 19:01 0:00 [ssl_client]
0 0 ? Z 19:02 0:00 [ssl_client]
0 0 ? Z 19:02 0:00 [ssl_client]
0 0 ? Z 19:02 0:00 [ssl_client]
0 0 ? Z 19:02 0:00 [ssl_client]
0 0 ? Z 19:03 0:00 [ssl_client]
0 0 ? Z 19:03 0:00 [ssl_client]
0 0 ? Z 19:04 0:00 [ssl_client]
0 0 ? Z 19:04 0:00 [ssl_client]
0 0 ? Z 19:04 0:00 [ssl_client]
0 0 ? Z 19:04 0:00 [ssl_client]
0 0 ? Z 19:05 0:00 [ssl_client]
0 0 ? Z 19:05 0:00 [ssl_client]
0 0 ? Z 19:05 0:00 [ssl_client]
0 0 ? Z 19:05 0:00 [ssl_client]
0 0 ? Z 19:06 0:00 [ssl_client]
0 0 ? Z 19:06 0:00 [ssl_client]
0 0 ? Z 19:07 0:00 [ssl_client]

all above zombie process was found after upgrading the edge container at 19:01 by watchtower.

Thank you!

EugeneOne1 · 2023-03-31T13:49:10Z

@leo15dev, what exact version are you using? We've investigated the case and have some assumptions. The Docker container runs the command specified by the ENTRYPOINT under PID=1. In UNIX systems this exact process is intended to handle orphan processes (and zombies too) and is usually occupied by some init system. In case of Docker container, this PID is occupied by AdGuard Home itself, so these processes are becoming its children and stay in the process table forever due to lack of appropriate handling.

However, we've only been able to reproduce this "process zombification" requesting the web UI by wget with HTTPS scheme which triggered the ssl_client. We've been using the HTTPS scheme for these requests in the initial version of our healthcheck script, but the current implementation avoids it, so it's not really clear what calls ssl_client now.

leo15dev · 2023-03-31T14:30:39Z

@leo15dev, what exact version are you using? We've investigated the case and have some assumptions. The Docker container runs the command specified by the ENTRYPOINT under PID=1. In UNIX systems this exact process is intended to handle orphan processes (and zombies too) and is usually occupied by some init system. In case of Docker container, this PID is occupied by AdGuard Home itself, so these processes are becoming its children and stay in the process table forever due to lack of appropriate handling.

However, we've only been able to reproduce this "process zombification" requesting the web UI by wget with HTTPS scheme which triggered the ssl_client. We've been using the HTTPS scheme for these requests in the initial version of our healthcheck script, but the current implementation avoids it, so it's not really clear what calls ssl_client now.

AdGuard Home, version v0.108.0-a.493+2eb3bf6e

I run it behind the nginx reverse proxy. But in Encryption settings -> HTTPS port , I set it to 443. Because I assigned an independent docker network for AdGuard Home container. And in docker-compose.yaml, I do not expose any port.
Maybe this is the reason why produced the ssl_client?

tls:
enabled: true
server_name:
force_https: true
port_https: 443
port_dns_over_tls: 853
port_dns_over_quic: 853
port_dnscrypt: 0
dnscrypt_config_file: ""
allow_unencrypted_doh: false
certificate_chain: ""
private_key: ""
certificate_path: /opt/adguardhome/conf/ssl/cert.pem
private_key_path: /opt/adguardhome/conf/ssl/key.pem
strict_sni_check: false

Thak you!

EugeneOne1 · 2023-03-31T14:49:36Z

@leo15dev, well, it shouldn't be the case, since the health check script runs inside the same container it checks.

Since the v0.108.0-a.493+2eb3bf6e is 1 revision ahead of the one, which started avoiding HTTPS, I should ensure, if the issue still relevant? If these zombies still present after the update?

leo15dev · 2023-03-31T15:00:13Z

Since the v0.108.0-a.493+2eb3bf6e is 1 revision ahead of the one, which started avoiding HTTPS, I should ensure, if the issue still relevant? If these zombies still present after the update?

still producd the the zombies process.

2023/03/31 19:00:43.655326 [info] AdGuard Home, version v0.108.0-a.493+2eb3bf6e

root 2495058 0.0 0.0 0 0 ? Z 19:01 0:00 [ssl_client]
root 2495059 0.0 0.0 0 0 ? Z 19:01 0:00 [ssl_client]
root 2495509 0.0 0.0 0 0 ? Z 19:01 0:00 [ssl_client]
root 2495510 0.0 0.0 0 0 ? Z 19:01 0:00 [ssl_client]
root 2495974 0.0 0.0 0 0 ? Z 19:02 0:00 [ssl_client]
root 2495990 0.0 0.0 0 0 ? Z 19:02 0:00 [ssl_client]
root 2496455 0.0 0.0 0 0 ? Z 19:02 0:00 [ssl_client]
root 2496456 0.0 0.0 0 0 ? Z 19:02 0:00 [ssl_client]
root 2496895 0.0 0.0 0 0 ? Z 19:03 0:00 [ssl_client]
root 2496896 0.0 0.0 0 0 ? Z 19:03 0:00 [ssl_client]
root 2497386 0.0 0.0 0 0 ? Z 19:04 0:00 [ssl_client]
root 2497387 0.0 0.0 0 0 ? Z 19:04 0:00 [ssl_client]

Same in

2023/03/31 23:00:37.289831 [info] AdGuard Home, version v0.108.0-a.494+1731ce9c

root 2726031 0.0 0.0 0 0 ? Z 23:01 0:00 [ssl_client]
root 2726046 0.0 0.0 0 0 ? Z 23:01 0:00 [ssl_client]
root 2726798 0.1 0.0 0 0 ? Z 23:01 0:00 [ssl_client]
root 2726799 0.1 0.0 0 0 ? Z 23:01 0:00 [ssl_client]
root 2727385 0.2 0.0 0 0 ? Z 23:02 0:00 [ssl_client]
root 2727386 0.2 0.0 0 0 ? Z 23:02 0:00 [ssl_client]

Thank you!

Merge in DNS/adguard-home from 3290-kill-zombies to master Updates #3290. Squashed commit of the following: commit 3e06260 Merge: 5aa7aa4 1731ce9 Author: Eugene Burkov <[email protected]> Date: Fri Mar 31 20:04:04 2023 +0500 Merge branch 'master' into 3290-kill-zombies commit 5aa7aa4 Author: Eugene Burkov <[email protected]> Date: Fri Mar 31 16:38:00 2023 +0500 docker: add doc commit 52a0b67 Author: Eugene Burkov <[email protected]> Date: Fri Mar 31 14:41:41 2023 +0500 docker: add init emulator

leo15dev · 2023-03-31T15:50:10Z

I think AdGuard Home, version v0.108.0-a.495+f191cb07 fixed it, thank you!

EugeneOne1 · 2023-03-31T15:56:16Z

@leo15dev, great, thanks for your help. I suppose, the issue can be closed for now. Please feel free to open new ones for faced problems.

agneevX · 2023-04-15T12:45:48Z

I sort of regret creating this issue as I've learnt over time that Docker healthchecks create unexpected issues and do more harm than good.

I therefore ask that this feature be removed or that docs be updated to mention issues with this and how it can interfere with something as critical as DNS.

chennin · 2023-08-21T20:14:24Z

What in the world is the random string of letters in the ?dns= query and how do I construct it myself?

This is apparently hard to find an answer to, but I found your question via Google when I had the same question, so here's the answer for anyone who needs it.

It seems Adguard doesn't take a name/type parameter. But here's how to craft the dns parameter it does take in one Python line. Note: requires dnspython >= 2.2.0, or for you to remove the ,id=0 piece. Change example.com as you wish.

    DOMAIN="example.com" python3 -c "import os, dns.message, requests, base64; req = dns.message.make_query(os.environ['DOMAIN'], 'A', id=0).to_wire(); print(base64.urlsafe_b64encode(req).decode('utf-8').rstrip('='))"

Note that the response will be binary, so do not use with curl and output to a terminal. Interpreting the response can also be done in Python, example here (use requests to get the response then operate on r.content).

(reference re padding and id: here and on)

ainar-g added the feature request label Jun 28, 2021

ainar-g mentioned this issue May 19, 2022

Docker container improvements #4575

Closed

ainar-g assigned EugeneOne1 Sep 7, 2022

ainar-g added the P3: Medium label Sep 7, 2022

ainar-g added this to the v0.107.13 milestone Sep 7, 2022

ainar-g modified the milestones: v0.107.13, v0.107.14 Sep 14, 2022

ainar-g modified the milestones: v0.107.14, v0.107.15 Sep 29, 2022

EugeneOne1 modified the milestones: v0.107.15, v0.107.16 Oct 3, 2022

EugeneOne1 mentioned this issue Mar 28, 2023

Consider custom port in Dockerfile healthcheck #5645

Closed

3 tasks

EugeneOne1 closed this as completed Mar 31, 2023

ainar-g mentioned this issue Apr 3, 2023

Server began to freeze after the last update #5661

Closed

3 tasks

ainar-g modified the milestones: v0.107.29, v0.107.28 Apr 12, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Dockerfile does not have health check #3290

Dockerfile does not have health check #3290

agneevX commented Jun 28, 2021

agneevX commented Jun 28, 2021

ainar-g commented Jun 28, 2021

sachinwadhwa commented Jun 28, 2021

ameshkov commented Jul 5, 2021

agneevX commented Jul 12, 2021

lolgast1987 commented Aug 3, 2021

PeterDaveHello commented Sep 7, 2022

PeterDaveHello commented Sep 7, 2022

JaneJeon commented Feb 13, 2023

PeterDaveHello commented Feb 14, 2023

JaneJeon commented Feb 14, 2023

PeterDaveHello commented Feb 14, 2023

zoonderkins commented Feb 23, 2023

alexsannikov commented Mar 9, 2023

EugeneOne1 commented Mar 27, 2023 •

edited

Loading

stavros-k commented Mar 27, 2023

EugeneOne1 commented Mar 28, 2023

stavros-k commented Mar 28, 2023

leo15dev commented Mar 29, 2023

EugeneOne1 commented Mar 31, 2023

leo15dev commented Mar 31, 2023

EugeneOne1 commented Mar 31, 2023 •

edited

Loading

leo15dev commented Mar 31, 2023 •

edited

Loading

leo15dev commented Mar 31, 2023

EugeneOne1 commented Mar 31, 2023

agneevX commented Apr 15, 2023

chennin commented Aug 21, 2023 •

edited

Loading

Dockerfile does not have health check #3290

Dockerfile does not have health check #3290

Comments

agneevX commented Jun 28, 2021

agneevX commented Jun 28, 2021

ainar-g commented Jun 28, 2021

sachinwadhwa commented Jun 28, 2021

ameshkov commented Jul 5, 2021

agneevX commented Jul 12, 2021

lolgast1987 commented Aug 3, 2021

PeterDaveHello commented Sep 7, 2022

PeterDaveHello commented Sep 7, 2022

JaneJeon commented Feb 13, 2023

PeterDaveHello commented Feb 14, 2023

JaneJeon commented Feb 14, 2023

PeterDaveHello commented Feb 14, 2023

zoonderkins commented Feb 23, 2023

alexsannikov commented Mar 9, 2023

EugeneOne1 commented Mar 27, 2023 • edited Loading

stavros-k commented Mar 27, 2023

EugeneOne1 commented Mar 28, 2023

stavros-k commented Mar 28, 2023

leo15dev commented Mar 29, 2023

EugeneOne1 commented Mar 31, 2023

leo15dev commented Mar 31, 2023

EugeneOne1 commented Mar 31, 2023 • edited Loading

leo15dev commented Mar 31, 2023 • edited Loading

leo15dev commented Mar 31, 2023

EugeneOne1 commented Mar 31, 2023

agneevX commented Apr 15, 2023

chennin commented Aug 21, 2023 • edited Loading

EugeneOne1 commented Mar 27, 2023 •

edited

Loading

EugeneOne1 commented Mar 31, 2023 •

edited

Loading

leo15dev commented Mar 31, 2023 •

edited

Loading

chennin commented Aug 21, 2023 •

edited

Loading