Gitaly North Star Metric 🌟

📖 Background and Overview

• North Star Metrics Presentation
• Deck from North Star Metrics Presentation

▶ Define your product feature's key benefit/value to customers

Gitaly is a critical service that provides read and write access to Git repositories for:

• users, through the GitLab application, SSH (via shell), and HTTP (via workhorse)
• continuous integrations services through the Git protocol to GitLab runners and other CI services

Loss of Git access, or service degradation prevents developers from being productive, and prevents businesses from deploying features and fixes to their customers.

🥅 Define north star metric that quantifies delivery of this "must have" value

Monthly Gitaly availability.

Gitaly is defined as available if:

• Latency Apdex score is above 0.95
• Error rate is below 0.10%

This is aligned with our GitLab.com availability metrics.

• This supports the revenue growth of:
  • GitLab.com because service availability is critical for making it enterprise ready
  • Self hosted because the availability and performance are critical evaluation criteria
• This reflects the customer getting value because Git it ensure the core Git service is reliable and fast
• This measures progress because as we improve Gitaly availability should increase by.
  • Improving performance reduces latency
  • Improving quality reduces error rates

❗ What do you need from the Telemetry/Data team?

• The data is available but I need analytics to visualize/re-structure
• The data is not available and I need to implement tracking
• I don't know, need help to check whether this data exists or not

Latency and error rate metrics are available from Gitaly via Prometheus, and are available in Grafana.

• For GitLab.com, these metrics are not being added to the data warehouse
• For self hosted customers, these metrics are not available through the usage ping

🧩 Answer these questions to provide more context

• Any alternative north star metrics you would consider, either because they are easier to get, or you are debating between that and the proposed NSM?
• Does your feature play a bigger role in initial purchase, retention, or expansion/upgrade?
• Does your feature play a bigger role in SMB sales, Enterprise sales, or both?
• What are some dimensions you would expect to slice/dice your north star metric? For example, by plan tier, by free/paid, etc.
• Are there any GitLab product features that are "downstream" or "upstream" to your product feature? For example, customers have to use feature A to use your feature, or once customers use your feature, they are more likely to use another feature.
• Please attach a mock-up, drawing, or chart in another system that will think will allow you to make a data-driven decision for the propose metric. Link: ______
  • Please provide an example of how you would use the mock-up or chart to make the data-driven decision.

📈 Define Supporting KPI/metrics

• Inputs:
  • Latency, this is reported via Prometheus
  • Error rate, this is reported via Prometheus

Improving these metrics directly improves availability.