Add `dig` binary to image to allow defining health probes about dnscrypt-proxy-docker HOT 11 CLOSED

from dnscrypt-proxy-docker.

Wow, thank you for this! I've finally added a healthcheck to my last container.

Would be great if you could add this info to docker hub & repo readme (edit: just noted this info is already on repo readme).

from dnscrypt-proxy-docker.

Good catch @CHerSun , I found a bug in the script that is supposed to update the DockerHub description and it should be fixed next time we merge a PR.

from dnscrypt-proxy-docker.

My only requirements are:

• Being able to query on non-standard ports (i.e. not only 53)
• Have the command exit with non-zero code if the query fails

I've done a quick check and seems like doggo can do the former @udp://10.0.0.5:5533, but I haven't managed to achieve the latter

roobre@Archiroo
12:31:06   ~ $> doggo foo.bar.local A @udp://1.1.1.1:53
NAME	TYPE	CLASS	TTL   	ADDRESS                       	NAMESERVER	STATUS
.   	SOA 	IN   	86400s	a.root-servers.net.           	1.1.1.1:53	NXDOMAIN	
    	    	     	      	nstld.verisign-grs.com.       	          	        	
    	    	     	      	2023032300 1800 900 604800    	          	        	
    	    	     	      	86400                         	          	        	

roobre@Archiroo
12:31:13   ~ $> echo $?
0

Taking a look right now to see if there's any cli flag I'm missing.

from dnscrypt-proxy-docker.

Seems like there's a bug in doggo which prevents it from existing with non-zero code even when it should. I've reported it here mr-karan/doggo#68

from dnscrypt-proxy-docker.

After further investigation, it seems like an NXDOMAIN also does not cause doggo to exit with an error, this time it seems by design. This would make hard to use it to use it as a probe, especially if the image does not contain grep or a shell to hack around this limitation :(

from dnscrypt-proxy-docker.

I think the NXDOMAIN limitation is true of dig as well: https://linux.die.net/man/1/dig

Dig return codes are:
0: Everything went well, including things like NXDOMAIN
1: Usage error
8: Couldn't open batch file
9: No reply from server
10: Internal error

from dnscrypt-proxy-docker.

:oof:, that certainly complicates things 🤔
Not sure what the best approach could be here. Took a look at the PR linked above, and I think that it would work, but seems like a lot of complexity for a simple probe.

Perhaps we could do an ad-hoc go program for this, WDYT? Compiling it may be annoying but I think it can be less annoying than copying dynamically linked binaries from alpine.

I wouldn't mind hacking something together on a PR if you're okay with this :)

from dnscrypt-proxy-docker.

By all means, PRs are welcome!

from dnscrypt-proxy-docker.

@klutchell Gave this a swing in #117, LMK what you think!

from dnscrypt-proxy-docker.

Resolved by #117

from dnscrypt-proxy-docker.