basecamp / easymon Goto Github PK

I have a site where the following path is being picked up by easymon:

/uppababy-2018-cruz-stroller-emmett-green.html

In the site, this path is a 404, which should then get picked up by our error controller and redirected to a configured destination but it seems like easymon is picking it up before that.

{"method":"GET","path":"/up/pababy-2018-cruz-stroller-emmett-green.html","format":"html","controller":"easymon/checks","action":"show","status":404,"duration":0.92,"view":0.49,"ip":"72.94.148.85","route":"easymon/checks#show","request_id":"7ad5a34f61be33b45a268c486ccd7e0d","source":"app-7f76cd6859-8z497","tags":["request"],"@timestamp":"2019-05-17T14:37:57.568Z","@version":"1"}

It is unclear how the path is changing from /uppababy-2018-cruz-stroller-emmett-green.html to /up/pababy-2018-cruz-stroller-emmett-green.htm

Background: Rails 7.1 will be adding health checks. I work for Fly.io as a Rails Speciality and would like to make Fly.io automatically detect the presence of health checks and pre-configure applications to enable these health checks if I can detect them. The problem is that health checks can vary on paths and implementation. I have a pull request that has been merged into Rails 7.1, and a pull request open for the rails_healthcheck gem that enable me to discover the healthcheck endoint using the following code:

bin/rails runner "p Rails.application.routes.url_helpers.rails_health_check_path"

Easymon poses unique challenges. For starters, there can be multiple endpoints, and it would be desirable to provide a way to indicate which one is the default one. Easymon is also implemented as an engine, so the url helper is scoped to the engine.

Rather than making assumptions, I'm opening up an issue so we can discuss how to approach this problem. If there is code I can contribute, I'm willing to do the work. If there is a recommendation on how I can adjust my check, let me know.

See also:

• superfly/flyctl#1678
• rails/rails#47217
• rails/rails#46936

We are experiencing issues addressed with the following commit:
9472135#diff-5892261b17dbd431a67ac4302cb920ca

Wondering if you would be able to release a new version to ruby gems when you have a moment.

Thanks.

Currently, it's a huge float value in seconds. This should be in milliseconds, and rounded to 3 decimal places (because arbitrary).

When passing an invalid check name to ChecksController#show, an exception is raised, resulting in a 406 response with the following body:

"No check named '#{name}'"

Firefox renders that body, and if the name includes any JavaScript code via <script> tags, it'll be executed.

The current ActiveRecordCheck fails on Postgres, because SELECT 1=1 returns true, not 1 as the check expects.

On Postgres:

postgresql=> SELECT 1=1;
 ?column? 
----------
 t
(1 row)

postgresql=> SELECT 1;
 ?column? 
----------
        1
(1 row)

On MySQL:

mysql> SELECT 1=1;
+-----+
| 1=1 |
+-----+
|   1 |
+-----+

1 row in set (0.00 sec)
mysql> SELECT 1;
+---+
| 1 |
+---+
| 1 |
+---+
1 row in set (0.00 sec)

Hence the check at https://github.com/basecamp/easymon/blob/master/lib/easymon/checks/active_record_check.rb#L21 fails on Postgres.

Would you consider changing the check to just SELECT 1, which would work on MySQL, Postgres, SQLite, and probably quite a few others?
Should I submit a pull request for the change?

The idea being to return a richer data structure when we have a critical checklist. The critical checklist is getting flattened into a string:

"critical": {
        "success": true,
        "timing": "2.811",
        "message": "memcached: Up - 0.747ms\nthird-database: Up - 0.445ms\nother-database: Up - 1.214ms\ndatabase: Up - 0.271ms\n - Total Time - 2.677ms"
    }

This isn't ideal.