Skip to main content

Smarter retries strategies: same region and backoff

Feature RequestScheduling

The current retry strategy for all checks is fairly limited. We want to expand the possibilities for retrying with two options:

- same region: right now, retries are always scheduled on a different region than the original attempt. This is not ideal for geo-fenced checks or region-aware websites rendering different content for different regions.

- backoff strategies: we want to enable multiple retries according to a strategy the user can define: linear or exponential so the target system has more time to recover from an initial failure.

Posted by Tim Nolet ·
Status: Completed6 comments

Log in to comment and vote

Comments6

  • Tim Nolet

    •

    Sep 5, 2023

    Hi folks, as you might have seen, we shipped this item last Friday.

  • Tim Nolet

    •

    Aug 28, 2023

    Merging this ticket to https://checkly.featurebase.app/submissions/64cbb6e83da95c1941702936

  • Tom Elliott

    •

    Aug 7, 2023

    A little context from our usage: we turned off retries for most of our checks because users would become confused when they saw the last random region that was used for a retry. Being able to select a subset of regions could be useful to us in bypassing region-specific issues but also still having the region shown in a failure "make sense".

    We ended up setting our checks to alert after 4 consecutive failures, so any improvements that could allow us to get faster notification of downtime reliably would be amazing 😃

    • Tim Nolet

      •

      Aug 9, 2023

      Hey Tom, these are valuable insights. We are speccing and prototyping as we speak and starting work soon on this.

  • Pete

    •

    Jun 8, 2023

    This would be extremely useful for our Checkly use-case as we have quite some time between checks and often experience transient errors, where the current "double check" occurs too soon after the initial check.

    • Tim Nolet

      •

      Jun 8, 2023

      He Pete, we will solve this going forward with backed off retries. This means we do not fuss with the static configuration of your check, but allow you to retry in a sane way so in between let's say a 1hr check schedule, you get multiple retries.

      We are rounding of some other work first and then dive into this.