Werk #11095: AWS S3

This werk fixes multiple issues with the AWS S3 monitoring, i.e.:

• As a follow-up for werk #11092, where one service per monitored region for the monitoring of AWS limits was introduced, the corresponding service for S3 is now always discovered as 'AWS/S3 Limits Global'. This is because there are no per-region limits for S3, only global account-wide limits. Users have to re-discover this service and potentially adjust any related rules for this to come into effect.

• The regions set when configuring the AWS special are now properly taken into account when monitoring S3 buckets (note that they are deliberately ignored for the limits). This means that buckets residing in regions which are not monitored are not discovered any more and do not contribute to the S3 summary service. Users have to re-discover the S3 services for this to come into effect.

• AWS S3 CloudWatch metrics are now fetched properly. Previously, only the non-request metrics for buckets in the region us-east-1 worked, such that for example bucket sizes for buckets outside us-east-1 were wrong.

• A couple of metrics and some units were fixed.

• The rule "AWS/S3 HTTP Errors" for the check aws_s3_requests.http_errors has been added.

• The levels for the check aws_s3_requests.latency (rule "AWS/S3 Latency") are now specified in ms, before, the smallest possible level was 1 s. After installing the werk, users have to re-define this rule.

• Rules configured for the check aws_s3_requests ("AWS/S3 Bucket Requests") now have an effect, previously, they were ignored.

• Services based on S3 CloudWatch metrics now go stale instead of UNKN if AWS stops delivering data, for example in the case where requests to an S3 buckets stop.

• Names and tags set when configuring the AWS special agent are now handled correctly and excluded buckets are not fetched any more (but they are still counted for the limits).

To the list of all Werks