gusto/aptible_datadog_postgres_monitor

A docker container for Aptible containing the data dog agent to watch and provide metrics on Postgres instances.

aptible_datadog_postgres_monitor's Issues

Disable default metric collection

There are a bunch of default metrics that the agent tries to collect that aren't relevant for collecting postgres metrics from Aptible. Many of these can apparently be turned off by removing the .default files in /etc/dd-agent/conf.d

For example:

root@deadbeef:/etc/dd-agent/conf.d# ls -l *.default
-rw-r--r-- 1 root root  525 Feb 13 19:30 agent_metrics.yaml.default
-rw-r--r-- 1 root root 2118 Feb 13 19:30 disk.yaml.default
-rw-r--r-- 1 root root  790 Feb 13 19:30 network.yaml.default
-rw-r--r-- 1 root root  329 Feb 13 19:30 ntp.yaml.default

Not a major concern, but just annoying to see irrelevant failures in the logs like:

IOError: [Errno 2] No such file or directory: '/host/proc/filesystems'

Are you seeing all the metrics?

Hi there "Gusto"!
I just have a question: when you booted up your container in Aptible, did the dd-agent report all the metrics as is seen here: https://datadog-prod.imgix.net/img/blog/postgresql-monitoring/postgresql-monitoring-dashboard-pt1v2.png?fit=max

I'm only seeing a small subset of these metrics being reported.
Any info and advice is greatly appreciated!
M

Use specific Dockerfile.development file for local development

The dockerfile CMD is only used if you're running the docker image locally not in interactive mode. Similarly, the postgres package isn't used in production. Might want to consider at some point having a Dockerfile.development to reflect this.

Look @ Flake8 and apply to entrypoint.py

Some things it will probably bring up:

1. Remove re
2. Make sure you understand where yaml is coming from (IIRC it's not a python stdlib thing)
3. % should be replaced by .format
4. don't need your own wrapper function for retrieving a key / value pair, see default values on .get:

>>> os.environ.get('BLASH_1', 'woof')
'woof'

5. line 38 - can replace my_cond != None: with my_cond:
6. Sort imports by alphabetically. Just doing alphabetically is fine for now. If you really want to be rigorous, https://github.com/PyCQA/flake8-import-order
7. Use a if __name__ == '__main__': instead of just executing the script
8. Should the shebang line be #!/usr/bin/env bash ? Dennis will be a better authority here

Lock & use specific Python version in dockerfile

If using 2.7, lock that in the dockerfile and call with it explicitly. if already explicitly version locked in root dockerfile, should be good to go.

SSL mode hard-coded to false, prevents connection

I'm not sure if this is peculiar to the particular Postgres instances I'm looking at, but on my Aptible account, I cannot connect to the databases unless the ssl flag is set to true. The error I receive is:

InterfaceError: md5 password authentication failed

When I log into the built container (using something like aptible ssh) I can see configs are being properly written out to /etc/dd-agent/conf.d/postgres.yaml like:

- dbname: postgres
  host: db-blah-1234.aptible.in
  password: foo
  port: 12345
  ssl: false
  tags:
  - aptible_env:foo
  username: datadog

And when I test these configs, using python /opt/datadog-agent/agent/agent.py check postgres --profile in that container, I receive connection errors until I set the SSL flag to true (at which point everything works nicely again). It's currently hard-coded to false at

aptible_datadog_postgres_monitor/entrypoint.py

Line 18 in 0fdf3e0

'ssl': False,

Perhaps Aptible have turned on SSL support for their database instances recently?