Weekday morning Amazon SQS queue health digest in Slack

Every weekday at 8am, run a morning Amazon SQS health digest and post it to a Slack channel the user picks at setup. The trigger is a cron schedule the user can adjust (day of week, hour, timezone).

Start by calling Amazon SQS List Queues to enumerate every queue in the configured AWS region. If a queue-name prefix is provided at setup, pass it as the filter; otherwise list everything in the region.

For each queue URL returned, call Amazon SQS Get Queue Attributes and read ApproximateNumberOfMessages (visible backlog), ApproximateNumberOfMessagesNotVisible (in-flight messages consumers are working on), ApproximateAgeOfOldestMessage (how long the oldest message has been waiting), and RedrivePolicy (whether the queue targets a dead-letter queue). Use RedrivePolicy together with the queue name to tell whether the queue is itself a dead-letter queue or has one configured downstream.

Classify each queue as healthy, backing up, or DLQ-populated. A queue is DLQ-populated when it looks like a dead-letter queue (name suffix like -dlq or -deadletter, or it is referenced as the deadLetterTargetArn in another queue's RedrivePolicy) and it currently has any messages in it. A queue is backing up when its visible message count or its oldest-message age crosses a threshold the user can tune (defaults: more than 1,000 visible messages, or an oldest message older than 15 minutes). Everything else is healthy.

Rank the concerning queues first: DLQ-populated queues at the top, then backing-up queues sorted by oldest-message age descending. Group healthy queues into a single compact roll-up at the end so on-call sees the concerning items immediately.

Post one grouped digest to the chosen Slack channel using Slack Send a Message. Use clear severity emoji at the start of each line (🚨 for DLQ-populated, ⚠️ for backing up, ✅ for the healthy roll-up), and for each concerning queue show the queue name, visible message count, in-flight count, and the age of the oldest message. Underneath each hot queue add one short plain-English sentence explaining what those numbers mean so an on-call engineer can act fast — for example, ‘12,400 messages waiting and the oldest is 42 minutes old, so consumers are draining slower than producers are publishing’, or ‘423 messages sitting in the dead-letter queue, meaning downstream processing has been failing repeatedly and needs investigation before retention expires’.

If List Queues returns no queues at all, still post a single friendly note to the Slack channel saying no queues were found in the region so silence is never ambiguous.