I was recently struggling with getting a CloudWatch alarm on a log filter metric to work. I was refactoring the creation from a CloudFormation template into just Lambda Python code. I was able to get both the filter metric (looking for errors and exceptions in logs) and alarm created.

But, when I tested the alarm, it never went into an alarm state, even when then metric filter showed active data. It was driving me crazy. I was able to manually create a new alarm, pointing to the same metric filter and see it work. This is the type of bug that developers hate: a completely illogical bug (as viewed by the bug’s creator).

Here is the metric showing data:

Here is what the empty alarm looked like:

Searching online for an answer was not easy. My heart sunk when Stack Overflow didn’t have the exact answer to my question with my first search. :) Searches for “alarms not working for filter metrics” weren’t leading anywhere.

I slept on it and tried different searches the next day. Luckily, Marta Tatiana blogged her solution once she finally found it: Insufficient data: CloudWatch alarm based on custom metric filter. It turns out that you can’t specify the period units (seconds, minutes, …) when creating the alarm programmatically.

Marta demonstrated one of the best reasons for blogging: showing other people your solution to a problem using your words. Your words may end up being better SEO for some lucky developer down the road.

• aws