lorenzoaiello / terraform-aws-rds-alarms Goto Github PK

I'd like to be able to set the treat_missing_data option for the alarms while using this module. For example, I'd like to set it to notBreaching, to avoid instances where no data is captured by AWS or it erroneously marks the the alarm as insufficient data and then reports when it transitions back to OK.

When using RDS autoscaling of storage you can set a min and max allocation. Setting "disk_free_storage_space_too_low_threshold" is hard because it depends on the currently allocated space which is dynamic.

I really want a metric to tell me when I'm approaching the max storage allocation to know when AWS won't save me anymore. If you enable enhanced monitoring you can use the fileSys.used metric to trigger on this and it'd be nice if it were an option to enable.

Maybe some new options could be added around this conditional on a bool to know whether enhanced monitoring is enabled.

https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_Monitoring-Available-OS-Metrics.html

The README file says:

Alarms Always Created (default values can be overridden):

CPU Utilization above 90%
Disk queue depth above 64
Disk space less than 10 GB
EBS Volume burst balance less than 100
Freeable memory below 256 MB
Swap usage above 256 MB
Anomalous connection count

But the terraform documentation does not provide any hints on how to override values in resources defined by sub-modules. How are these default values overriden?

I'm wondering if there's anybody else out there whose using this module and wished they could configure only a subset of these alarms. Perhaps passing an enabled = False boolean in for the various types of checks. Loosely I think it could be:

• cpu
• disk
• memory
• anomaly

These can then be used in the aws_cloudwatch_metric_alarm resources as the cpu_credit resource does for the count parameter.

As of right now I just define variables such that the alarm won't actually ever alarm, which isn't ideal. I'd be willing to fork and give this a try if there's interest. Or if there's another way to accomplish this please enlighten me.

Hello,

Is it possible to add an example showing the use of this module with an SNS topic?

Thanks!

Hey, I/we avidly use this module in some foundational components and would like to offer to maintain this module moving forwards @lorenzoaiello if you're keen. It looks like you don't hardly get a chance to review/merge/update/support this repo. So we'd be happy to take this on. Would you mind? Please reach out to me on my email on my profile to discuss further.

And yes, I realize I could fork this but I think for the folks that have forked/starred/used this repo it would be good for them to transparently-ish be able to continue using this module without manually changing to an fork. Would be happy to transfer ownership to us to our org https://github.com/DevOps-Nirvana/ and add you as an fellow manager of this repo, or to simply add us as a contributor and give us the ability to do releases (eg: setup some Github Actions to do this automatically upon release).

Thoughts?

So I used this module to provision database alarms as per example. It did not work - all alarms were stuck in Insufficient data state. I did raise a case to AWS support so they looked at my alarms and identified that database ID was used where Identifier need to be used. This is for RDS SQL Server at least.
Your example says:
db_instance_id = aws_db_instance.default.id
Should be:
db_instance_id = aws_db_instance.default.identifier
Please amend documentation accordingly.

This line:

resource "aws_cloudwatch_metric_alarm" "cpu_credit_balance_too_low" {
  count               = length(regexall("(t2|t3)", data.aws_db_instance.database.db_instance_class)) > 0 ? "1" : "0"

results in:

Error: Invalid count argument

  on .terraform/modules/mymod.rds-alarms/main.tf line 28, in resource "aws_cloudwatch_metric_alarm" "cpu_credit_balance_too_low":
  28:   count               = length(regexall("(t2|t3)", data.aws_db_instance.database.db_instance_class)) > 0 ? "1" : "0"

The "count" value depends on resource attributes that cannot be determined
until apply, so Terraform cannot predict how many instances will be created.
To work around this, use the -target argument to first apply only the
resources that the count depends on.

probably the instance-class should not be read from the data provider but passed via a regular variable.