check_http: Unchunking body fails when no body is present

[AiYoriAoshi]

Checks with -N / --no-body fail with 2.3.3

# /usr/local/nagios/libexec/check_http -I <IP> -4 -N -v
GET / HTTP/1.0
User-Agent: check_http/v2.3.3 (monitoring-plugins 2.3.3)
Connection: close


http://<IP>:80/ is 1460 characters
STATUS: HTTP/1.1 200 OK
**** HEADER ****
Cache-Control: must-revalidate,public
Last-Modified: Tue, 18 Apr 2023  08:44:27 GMT
Content-Type: text/html
Transfer-Encoding: chunked
Connection: Keep-Alive
**** CONTENT ****
  [[ skipped ]]
Found chunked content
Chunked content did not start with a number at all (Line: 1415)
HTTP UNKNOWN: Failed to unchunk message body

as mentioned in 1815

Adding !no_body to the if as suggested fixes the problem for me.

[mike1356]

Adding !no_body to the if at line 1281 fixes it for me as well.

# /opt/plugins/check_http -H <IP> -p <port> -u /api/v1/health -N -v
GET /api/v1/health HTTP/1.1
User-Agent: check_http/v2.3git (monitoring-plugins 2.3git)
Connection: close
Host: <IP>:<port>

http://<IP>:<port>/api/v1/health is 128 characters
STATUS: HTTP/1.1 200 OK
**** HEADER ****
Server: Monkey/1.7.0
Date: Thu, 27 Apr 2023 11:59:14 GMT
Connection: Close
Transfer-Encoding: chunked
**** CONTENT ****
  [[ skipped ]]
HTTP OK: HTTP/1.1 200 OK - 128 bytes in 0.002 second response time |time=0.002058s;;;0.000000;10.000000 size=128B;;;0;

[RincewindsHat]

I will take a look at it. Until then, you could use check_curl as a replacement.

[moschlar]

It seems to me that the server is sending:
Transfer-Encoding: chunked but no Content-Length: 0.

(The server being IIS...)

[trickert76]

The problem is, that the parameter -N (--no-body) removes any kind of response after the header, but this commit seven months ago only reads the body and doesnt check, if the parameter --no-body is present.

[trickert76]

So, this is not a bug from the HTTP server (in my case an nginx). Its a bug in the way the new code for chunked content is handling requests when --no-body is defined. There is already a fix documented but not merged.

[juliogonzalez]

@trickert76 your patch seems to fix the issue for me. Thanks!

Maybe you should prepare a PR against https://github.com/monitoring-plugins/monitoring-plugins?

#1806 suggest that maybe check_http will be removed and it's mention that maybe new features should not be affected, but bugfixing should be OK.

[waja]

This should be fixed with #1901, isn't it?

[RincewindsHat]

@waja I think it is, but I am not completely sure, that everything is handled correctly if the server sends an empty body and nobody is not chosen.

[waja]

@trickert76 your patch seems to fix the issue for me. Thanks!

Maybe you should prepare a PR against https://github.com/monitoring-plugins/monitoring-plugins?

#1806 suggest that maybe check_http will be removed and it's mention that maybe new features should not be affected, but bugfixing should be OK.

a6802bd is exactly what #1815 (comment) is suggesting. So if that's the only change that is needed, we don't need another PR.

Can anybody having this issue please test master (which has a6802bd merged) and verify if the issue is done? thanks!

[juliogonzalez]

@waja, I just built the code from master an uploaded check_http to my checkmk instance.

It works fine after forcing a first check, but since the problem was not happening all the time in the past (at least in my case), please give me some time to confirm if it's fixed.

[juliogonzalez]

@waja, after a few days, it works for me.

So FMPOV, the bug report can be closed as solved.

Thanks a lot!