Skip to content
Weinan Si edited this page Sep 10, 2019 · 1 revision

Welcome to the OSDroid wiki!

issue sentinel rules

site issue

For running workflows whose:

  • running period (newest error report timestamp - oldest error report timestamp) > 4h
  • probability of ACDC from newest prediction > 0.5

, two error reports:

  1. newest error report
  2. error report generated around 4h ago

, are compared to get the error count increase for each reported site.

Aggregate increases from all workflows, flagged site which has 500 error count increase as site issue.

workflow issue

For running workflows whose:

  • running period (newest error report timestamp - oldest error report timestamp) > 1d
  • probability of resubmit from newest prediction > 0.3

, flag those

  • fraction of prediction of resubmit ranking first over past 1 day > 75%
  • number of totalError from newest error report > 100
  • failure rate from newest error report > 50%

as workflow issue.

Clone this wiki locally