Exim integration gives no metric data

[cw-Widad]

Output of the info page

    ------------
      Instance ID: exim:c933403ba680af29 [OK]
      Configuration Source: file:/etc/datadog-agent/conf.d/exim.d/conf.yaml
      Total Runs: 2,942
      Metric Samples: Last Run: 0, Total: 0
      Events: Last Run: 0, Total: 0
      Service Checks: Last Run: 1, Total: 2,942
      Average Execution Time : 1ms
      Last Execution Date : 2023-11-30 13:13:41 CET / 2023-11-30 12:13:41 UTC (1701346421000)
      Last Successful Execution Date : 2023-11-30 13:13:41 CET / 2023-11-30 12:13:41 UTC (1701346421000)

Additional environment details (Operating System, Cloud provider, etc):
Ubuntu 22.04.1 on Google Cloud Platform

Describe the results you received:
No metrics received even when queue has messages.
Results of running exim -bp | exiqsumm is

Count  Volume  Oldest  Newest  Domain
-----  ------  ------  ------  ------

  680  1312KB     69h     19h  foobar.xyz
---------------------------------------------------------------
  680  1312KB     69h     19h  TOTAL

Log files in /var/log/datadog/agent.log with log level set as debug shows these results

2023-11-30 12:38:41 CET | CORE | DEBUG | (pkg/collector/python/check.go:88 in runCheck) | Running python check exim (version: '1.0.0', id: 'exim:c933403ba680af29')
2023-11-30 12:38:41 CET | CORE | DEBUG | (pkg/collector/python/datadog_agent.go:135 in LogMessage) | exim:c933403ba680af29 | (subprocess_output.py:55) | Running get_subprocess_output with cmd: ['exim -bp', '|', 'exiqsumm']
2023-11-30 12:38:41 CET | CORE | DEBUG | (pkg/collector/python/datadog_agent.go:135 in LogMessage) | exim:c933403ba680af29 | (subprocess_output.py:59) | get_subprocess_output returned (len(out): 0 ; len(err): 0 ; returncode: 0)

Describe the results you expected:
Expected results with a Count of 680 as shown in command line execution above and a value of 1312000 for volume.

Additional information you deem important (e.g. issue happens only occasionally):

Some debugging was done on the code and as far as I see, it is an bug in the code. The metric is created by a command called here. exim/datadog_checks/exim/check.py

• The command is input wrong in the function call, it should be given as command and options separately, i.e ['exim', '-bp'] vs ['exim -bp'] . Datadog documentation also confirms this here. writing-checks-that-run-command-line-programs

• The function call uses two commands piped together, while the documentation for the get_process_output clearly outlines that a piped command won't work here base/utils/subprocess_output.py

• There are no tests run to check if the exim -bp command works here test_exim.py

[cw-Widad]

@JeanFred

[JeanFred]

Hi @cw-Widad, thanks for the details bug report. I’m afraid that since #1326, we have stopped using exim at $dayJob so I won‘t be able to work on this nor test any fix :-/

Regarding your notes:

• right, indeed, the options should have been in an array
• Aoutch, right, that was quite the oversight on my part. Yes, I guess that should be replaced with something like

command = ['exim', '-bp', '|', 'exiqsumm']
output = subprocess.check_output(command, shell=True)

• Yes, indeed, I could not find a good way to have a proper functional test for that :) (Smth like a docker-compose setup with exim running etc.)

[cw-Widad]

Thanks for your reply @JeanFred

I am not sure if piped commands is the way to go here, since Datadog documentation suggests not using subprocess python module directly (From the documentaion here : Since the Python interpreter that runs the checks is embedded in the multi-threaded Go runtime, using the subprocess or multithreading modules from the Python standard library is not supported in Agent version 6 and later.)
Unfortunately this means that exiqsumm won't be usable at all because exiqsumm only works on piped inputs.

I have tested using the exim -bpc command for just getting the total messages in queue, I will try rewriting the code for exim.queue.count using this and create a PR.