Defining Service Level Objectives for Loggregator
summary
Shows a concrete SRE approach: a black‑box app CF Log Mon and Concourse pipelines measure log stream reliability, operators target 99% (one year goal) and already hit 99% over six months for Pivotal Web Services while Pivotal Tracker lags. Capacity tests find a single Doppler fails near ~30,000 envelopes/sec and the team recommends 10,000 envelopes per second per Doppler; syslog adapter capacity is ~500 drains each. Includes repos, monitoring dashboards and CI so the tooling is reproducible.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 1:22Agenda
- 2:29Improving Message Reliability
- 2:59Before Defining Service Level Objectives
- 6:34Monitoring Your Log Stream
- 7:45Blackbox Monitoring
- 10:43Whitebox Monitoring
- 12:20Syslog Drains
- 13:40Setting Service Level Objectives
- 16:47Capacity Planning
- 17:35Monitoring
- 18:21Loggregator Operator Guide
- 19:00Agile Approach
- 20:05Key Themes
- 21:25Feature Proposals
- 21:44Review
- 22:13Key Features
- 23:51Flexible Egress
- 24:38Selectors
- 26:10A New Simplified World
- 27:46Dev Segmented Firehose
- 29:59Feature Narratives
- 30:37Outro
related talks
32mTuning and Managing Loggregator, the Cloud Foundry Logging Subsystem
22mLook Ma, No Metrics: A Quest Towards Extracting SLOs
32mAddressing Delivery Reliability in Loggregator
33mHow and why we lowered our SLO
30mReliable Observability at Scale: Error Budgets for 1,000+
35mSRE 101
23mGoing from 30 to 30 Million SLOs
49mScaling Reliability: So You Want to Add a 9