Skip to content
Branch: master
Commits on May 22, 2019
  1. Mark verify-tech-team-infrastructure as owning this repo

    philandstuff committed May 22, 2019
    This will mean that:
     - @alphagov/verify-tech-team-infrastructure will be notified of any
       PRs on this repo
     - members of that team can use the [review-requested][] feature to
       see outstanding PRs in one place rather than visiting each repo
    There is an optional feature that would mean that only
    verify-tech-team-infrastructure can submit reviews on this repo.  I
    do *not* intend using this; my goal for CODEOWNERS is just to improve
    the visibility of code that teams ought to have their eyes on.
Commits on May 10, 2019
  1. Merge pull request #219 from alphagov/actually-match-the-right-load-b…

    philandstuff committed May 10, 2019
    Fix cloudwatch_exporter config
  2. Fix cloudwatch_exporter config

    philandstuff committed May 10, 2019
    This regex didn't actually match any loadbalancers, so i ended up
    saving us too much money
  3. Merge pull request #218 from alphagov/reduce-metrics-even-more

    philandstuff committed May 10, 2019
    Reduce cloudwatch metrics even more
  4. scrape cloudwatch_exporter less frequently

    philandstuff committed May 10, 2019
    We were scraping cloudwatch_exporter at the default interval of 15
    seconds, but we use the default cloudwatch_exporter period of 60
    seconds - meaning that we only get new metrics once a minute at most.
    By scraping cloudwatch_exporter 4 times less frequently, we will make
    4 times fewer API calls.  📉
  5. Reduce cloudwatch metrics even more

    philandstuff committed May 10, 2019
    Request only metrics for the ingress load balancer, not all the
    others (there are 8 more loadbalancer/targetgroup combinations, so
    this will reduce our metrics to 3/11 of their current amount)
  6. reduce the number of CloudWatch metrics we collect

    philandstuff committed May 10, 2019
    We're spending quite a bit on GetMetricsStatistics API calls.  We can
    reduce that by requesting fewer metrics.
    For this commit, I focussed on retaining only those metrics required
    to support existing alerts.
    We don't use the RequestCount or RequestCountPerTarget or
    TargetResponseTime metrics in any alerts, so they can go.
    We use the Target_[2345]xx_Count metrics, but our alerting rules
    aggregate away the AvailabilityZone label.  So we can just not request
    it in the first place and reduce our bill on these metrics by a factor
    of 3.
You can’t perform that action at this time.