#What happens when the automation HTTP request fails?

The Custom step (HTTP request) has a Reliability block with three choices: whether the automation stops or continues when the other system answers with an error, how many times to retry before giving up, and how long to wait for the answer.

By default a new step stops and records the failure — that contact's run shows up in the Executions tab with the reason, instead of ending as if it had worked.

Role: any member with edit access to the workspace.

#Before you start

• An automation with a Custom step (Next step menu → Integration section).

#Steps

1. On the automation canvas, click the HTTP Request step to open its settings.
2. Scroll to the Reliability block.
3. Under If the request fails, pick:
   • Stop and record the failure — that contact's run ends there and shows up in the Executions tab with the reason. The following steps do not run.
   • Continue to the next step — the automation carries on as if nothing happened.
4. Under Extra attempts, pick from Do not retry up to 3 extra attempts.
5. Under Timeout (seconds), enter 1 to 120. The default is 30.
6. Click Save.

On the step card, the canvas then shows whatever differs from the default — for example Stops on failure · 2 extra attempts · 60s timeout.

#What counts as a failure

#Which failures are retried

Extra attempts apply only to temporary failures:

• 429 (the destination asked us to slow down);
• 408 and 5xx errors from the destination;
• timeouts, refused connections and other network drops.

A 4xx error is never repeated: the request is the same and so would be the answer. A rejected address is not retried either, for the same reason.

Between attempts the automation waits longer each time — about 30 seconds, then 2 minutes, then 8 minutes. If the destination says when it will be ready through the Retry-After header, that time is honoured; when it asks for more than 15 minutes, the automation gives up instead of holding the contact.

The contact waits between attempts — the following steps only run once the request finishes, either with success or with the last failure.

#The other system will not silently receive the same thing twice

Every request that sends data (POST, PUT, PATCH, DELETE) carries two headers:

If your system stores the key and ignores a repeated one, an extra attempt never becomes a duplicate request.

#Automations that already existed stay as they were

Steps saved before this update stay on Continue to the next step, with no extra attempts and a 30-second timeout — nothing changes on its own. To make one stop on failure, open the step, change the choice and save.

#If it did not work

• The run does not show up as a failure: check that the step is on Stop and record the failure. On Continue to the next step the error is only recorded on the step, and the run ends normally.
• The automation is taking too long to move on: with extra attempts, the contact waits out the interval between them. Lower the attempts or the timeout.
• The destination got the request twice: it is ignoring X-Clickmax-Idempotency-Key. Store the key on your side and reject the repeat.
• The error says the address is not allowed: the URL must start with https:// and point to a public address. Internal addresses and http:// are rejected.

#Related

• How do I branch on the HTTP request result?
• How do I see which contacts failed with each automation error?
• How do I see the data and variables of an automation run?