Explain why one candidate change ranks above another

In Summary, compare the reasons under the candidate changes. Timing, deployment evidence and the change’s relationship to the symptom all affect the order.

Read the reasons in ranking order#

When the leads shown rank for different reasons, each lead lists its reasons under it in the order they rank it. In an alert investigation, each repository's last successful deploy into the environment after the symptom began and by the time the alert opened reads Last deploy before the alert opened and ranks above every other lead. It says what was delivered, not what was serving. Like the latest deploy before the symptom, it gives a redeployed commit's redeploy time and its first deploy time.

Every lead then has one timing reason, which ranks it before any other reason: Landed before the symptom, or Landed after the alert's metric met its condition in its place, then Changed around symptom onset for a configuration change whose recorded interval contains the symptom start, then Landed during the symptom, then Landing time unknown when Proofline recorded no time for the change, such as a reviewed commit that no deploy or merge names. Landed after the alert's metric met its condition marks a lead that landed before the symptom window but after the time the Symptom window row gives for the metric meeting the alert's condition, and its time reads, for example, "deployed 10m before the assumed onset". The metric had already crossed the alert's threshold when it landed, so it is a weaker candidate than its time suggests, although it keeps its place in the list. Landed during the symptom reads Landed before the alert's metric first counted, or Landed before the alert's metric first met its condition, when the lead landed after the window began but no later than the time the Symptom window row gives for that, and its time reads, for example, "deployed 1m before the alert's metric first counted". It ranks as any lead that landed during the symptom does.

Among leads with the same timing, Proofline compares these reasons in this order:

  • deployed here, observed in inventory, in the latest deploy before the symptom or in the last deploy before the alert opened: a successful deploy into the environment delivered the commit, Kian observed the configuration change in the environment's inventory, the merge was in its repository's last successful deploy before the symptom, or, in an alert investigation, the merge was in the last deploy before the alert opened and not in the last one before the symptom. When the merge itself landed after the time the Symptom window row gives for the metric meeting the alert's condition, in the latest deploy before the symptom reads in the latest deploy before the assumed onset instead. None of these outranks another; the reasons below decide between them. A change that Kian saw only as a Cloud Asset Inventory asset appearing, disappearing or being touched also reads observed in inventory, but ranks below them, because short-lived resources change in every collection. A configuration change in which Kian saw a resource created and deleted again before the symptom began, such as a scheduled job's temporary secret, ranks in that same lower place: the resource no longer existed when the symptom started.
  • deploy failed: Proofline recorded only failed deploys of the commit into the environment. The lead ranks below the reasons above and above a lead with no record of reaching the environment.
  • not deployed before the symptom: no successful deploy of its repository before the symptom contained the merge, so no deploy had delivered it when the symptom began. A merge that an earlier deploy delivered before a rollback does not get this reason, and neither does one when Proofline could not read its repository's deploys at some point between the merge and the symptom. In an alert investigation, a merge that the last deploy before the alert opened delivered reads in the last deploy before the alert opened instead. The lead ranks last among leads with the same timing. A commit whose only deploys failed can show both deploy failed and this reason; it then ranks as a failed deploy.
  • serious review finding: Proofline's review of the commit has an unresolved serious finding, or published an environment finding from Kian in this environment. Kian's findings carry no severity, so each counts as serious, including in the serious findings that the review's timeline entry shows.
  • latest deploy before the symptom: the commit is its repository's last successful deploy into the environment at or before the symptom start. When that deploy finished after the time the Symptom window row gives for the metric meeting the alert's condition, the reason reads latest deploy before the assumed onset instead. The lead ranks above every lead that the reasons above leave equal with it, configuration changes included. A commit deployed a day before the symptom therefore ranks above a configuration change Kian observed minutes before it. The reason says what was delivered, not what was serving. After a redeploy or a rollback, the lead gives two times: when the commit was redeployed, which is the deploy that made it the latest, and when it was first deployed.

Leads that still tie rank by how close they landed to the symptom start. In a question's investigation, pinned at intake marks the commit of a review that Proofline selected when you asked: from the pull request you asked on or, in Slack, from the environment's linked repositories. It does not move the lead.

Search help

Tab to a result. Enter to open. Escape to close.