Postfix queue sizes dd-agent check #610
Conversation
…nfig example) dd-agent check
| if not exists(queue_path): | ||
| raise Exception("PostfixQueuesCheck: (%s) queue directory does not exist" % queue_path) | ||
|
|
||
| metric_name = '.'.join(['postfix.queues', queue]) |
There was a problem hiding this comment.
In general, are there only three queues? If so, I would recommend using tags instead of a metric.
postfix.queue.size{queue:active}
postfix.queue.size{queue:incoming}
postfix.queue.size{queue:deferred}
This make it much easier to graph in Datadog. For example, it's very simple to graph the size of all queues in one graph statement vs. one for each queue. For example:
sum:postfix.queue.size{role:email-server} by {queue}
If there are thousands of different queues, it's probably not good to use tags, but if there is a known, fixed amount, tags are great.
There was a problem hiding this comment.
There are around 12 possible named queues in postfix. This has not changed very much since postfix's inception. The 3 i specify in the yaml config example are the only ones that will get any use by postfix daemons, generally speaking. So, it sounds like tags are the way to go.. Thanks!
There was a problem hiding this comment.
Cool. Thanks for looking.
|
Nice! Thanks for doing this. I made a few small comments, but all in all, great stuff. One other general comment ... is it possible to write a small unit test that sets up a fake postfix directory, counts the things in it, and makes sure we count the right things? |
|
Awesome.. thanks for the feedback. I needed this for work and am more of a Rubyist.. and my Python blows. I can tackle every thing you said here this weekend.. I did do some unit testing locally, i assume you'd like something included? |
| @@ -0,0 +1,51 @@ | |||
| from os import walk, geteuid, popen | |||
| from os.path import exists, join | |||
| from collections import namedtuple | |||
There was a problem hiding this comment.
namedtuple was introduced in python 2.6
For compatibility purposed can you import it from our util module instead ?
from util import namedtuple
|
finishing up the discussed changes tonight/tomorrow... |
ghost
assigned …istory etc.. for the changes i am going to commit in a few minutes.
|
Hey guys.. per @clutchski i changed the name of this check to just postfix (was postfix_queue). I think i took care of everything but the unit test here. I asked a couple of questions in the most recent commit of postfix.py (re: logging error instead of warning, etc..). I did my own testing and the results of this check are correct, but i will work on a unit test in the coming days. |
|
Ughh.. no time for unit tests yet.. going to try to crack something out in the next couple of days.. apologies :( |
|
This nose test is done. See commits for an official dd-agent/tests/test_postfix.py above (phantasm66@1ae3ac9). The results of the nose test: [jwebb@vagrant-centos64-x86_64 ~]$ nosetests --nocapture --tests=dd-agent/tests/test_postfix.py hostname: Unknown host User jwebb may run the following commands on this host: Test messges put into active = 2010 Test messages counted by dd-agent for maildrop = 1992 Ran 1 test in 1.070s OK |
…r postfix.py... i wrote the unit test :)
|
Bump* Any update on a merge with this? |
|
This is awesome. I will definitely use it if it gets merged in. |
|
Thanks @phantasm66 ! Merging it in. It will be part of the next release. |
Postfix queue sizes dd-agent check
|
Awesome... Thanks! |
A generic and configurable agent check for calculating individual postfix queue sizes (number of messages). Example YAML config included. Tried to stick with the standard way of writing a datadog agent check. Postfix queue permissions are strict so i had to make accommodations for different methods of accessing the queues.