Monitor and Security teams will be responsible for triaging all logging issues
Determine the long-term owner of the ultimate logging product and/or process
Monitor and Security teams will be responsible for defining logging standards
Clearly define how logging works at GitLab in the runbooks
Exit Criteria
Triaging logging issues using the TriageBot
Security team will triage security and compliance issues to either the security team or a dev team
For dev teams, the Monitor will triage issues and assign to appropriate dev team.
Triagebot is modified by the security team to send slack notifications if higher priority logging issues aren't resolved in a specified timeframe.
Monitor: APM group takes ownership of the overall logging products/process
Infrastructure will continue to own the system architecture and systems.
Monitor will slowly build out a comprehensive logging product to meet security and dev needs.
Logging Standards
Monitor will own logging standards for dev teams
Security will own logging standards for compliance purposes
Monitor and Security team will both release logging standards
Transfer all current state of logging to the logging (runbook)[https://gitlab.com/gitlab-com/runbooks/tree/master/logging/doc]
Include all known details of the current system (DELKE remains 3rd-party-only for now)
Describes what logs are logged where
Describes current and proposed log standards
Details how to hook up new analyzers
Details about how TriageBot works
Details about who owns what parts of the triaging process
Outcome
Security Automation is leading the logging effort at this time.
The DELKE project ticks a lot of the boxes that we wanted out of a logging system. To that end we are moving forward to make DELKE more DevOps ready and productizable.
We will move forward on Logging standards.
Other Investigations
What do other companies do?
Jobs and files in ParK in AWS (wife just started there)
Look into Presto (Facebook Distributed SQL)
HotWarm ElasticSearch
LabKit
LabKit is an application logging library Andrew Newdigate invented to help structure and standardize logging (similar to Graphite pings) throughout the Ruby and Go code bases
Where do logs go today?
Unstructured logs (redis, etc) sent to GCS and Stackdriver
Structured sent to ELK
Large structured logs sent to BigQuery (via Stackdriver)
As a GitLab employee, I know there is a single team I can talk to about all things logging at GitLab.
As a GitLab employee, I can find who is on the logging team by visiting a single handbook page.
As a GitLab employee, I can contact the logging team via dedicated slack channel.
As a GitLab employee, I can label GitLab issues for the logging team.
As a logging team member, I only have one set of logging services/infrastructure to manage and maintain.
As a paying, self-managed customer, I can easily set up a logging infrastructure for GitLab to log to.
Developers/Support
As a developer or support, I can easily find, search, view, and analyze logs up to 6 months old
As a developer or support, I can easily add a new logger without coordinating with any other team
As a developer or support, I can easily find one page in the handbook that details what services we log to for each type of data, and what the URLs are for their dashboards.
As a developer or support, I can easily find one page in the handbook that clearly specifies how to write log messages.
As a developer or support, I can easily add a log message with all of the required data.
As a developer or support, I can be confident that no private data is accidentally logged.
As a developer or support, I use the same logging library per language as every other developer at GitLab.
Infrastructure
As an infrastructure engineer, I can set up multiple sinks for different logs
As an infrastructure engineer, I can easily find, search, view, and analyze logs up to 1 week old
As an infrastructure engineer, I can easily specify the retention period for each data store and/or data type (i.e., security related)
Security/Compliance/BI
As a security researcher/security engineer/compliance staff, I can easily find, search, view, and analyze logs up to 12 months old
As a security researcher, I can easily correlate logs between sources and between GitLab subsystems.
As a security researcher, et al., I can easily use logs to train ML models with historic data.
As a security researcher, et al., I can easily use our logs to detect potential security issues before they become problems.
As a security researcher, et al., I can easily analyze the logs for sensitive or private data.
Git is a trademark of Software Freedom Conservancy and our use of 'GitLab' is under license
