Skip to main content

Monitor integrations every day

Outcome. Every day you know whether the integrations are healthy, and every item that needs a person is explained or handed to someone who can recover it. Once a week you confirm that the background work ran and that your thresholds and alerts still fit. Who takes part: Integration operator, Integration administrator. Stages that use standard pages (Process automations, Batch jobs, the standard Business events catalog) need System administrator or the access named on the role page; the Lattice roles do not open them. How often: every day (stages 1 to 4) and once a week (stages 5 and 6).

Stages 1 to 3 are read-only and open to the Integration operator role. The actions that change a failed item need the Integration administrator role, which is why stage 4 is a hand-over.

1. Read the workspace (Integration operator)​

The nine tiles of the Integration operations workspace summarize the whole environment on one screen, and each tile opens the list behind it, already filtered.

Done when: every tile is zero, or a number you already know and that is falling; the Inbound orders on hold tile counts only orders still waiting on their price hold.

2. Check the queues (Integration operator)​

The tiles say how many items wait; the queues say whether it is getting worse. Read the dead-letter queue, the recurring data jobs, the held orders and the failed business events.

Done when: the parked count and the oldest parked age on Queue status are no higher than at your last check, and Needs attention only on Recurring integration health lists no job you cannot explain.

3. Check calls and files (Integration operator)​

Yesterday's statistics show failure and slowdown trends that single failed calls hide, and the inquiry shows whether partner files arrived and were processed.

Done when: yesterday's Failures and Average duration (ms) on Service call statistics look like a normal day for each operation, and Inbound message inquiry shows recent Received times on Bundles and records reaching Processed on Segments.

4. Hand over what needs recovery (Integration operator, then Integration administrator)​

You see every failed item, but retry, hold, release, discard, fix and resubmit, closing a claim and repairing a recurring data job belong to the Integration administrator role. Pass each item you cannot explain to an administrator with what View payload and the logs showed, and work it through the failure process.

Done when: at the next daily check, the items you handed over are gone from Dead letters awaiting action, Stale idempotency claims, Claims needing manual reconciliation and Recurring jobs failing or broken, or those counts are falling.

5. Weekly: confirm the background work ran (Integration operator, System administrator)​

Most of what you monitor is produced by Lattice's background processes, and a stopped process looks like a quiet day. The standard pages for this need access the Lattice roles do not grant, so a System administrator may check for you.

Done when: View most recent results shows recent runs for the processes that had work, Service call statistics has rows for yesterday, no dead letter stays Retrying long after its Next retry, and no file record stays Queued and no order stays Pending.

Compare the week with the weeks before, and adjust what counts as unhealthy so that the tiles and alerts flag real problems and nothing else.

Done when: every tile that is not zero has a known cause, each recurring data job's thresholds fit how often it receives files, and the alert events you rely on are active for the legal entities you need.

If something goes wrong​