Security Logging
From 1.3, OLIVE emits a defined catalogue of security events with stable codes, in a machine-readable form, so that alert rules can be written against codes rather than against log message text that changes between releases.
The full catalogue and how to alert on it is in Log Events and Automated Reactions. It sits in the Cluster Operator guide because they need to know what the platform emits in order to size, route and retain it — but the alert rules written against these codes are yours, along with first-line triage of what they fire. Your cluster operator provides the pipeline and the rule engine; the judgement about what an OLIVE event means is not theirs to make.
OLIVE writes logs and labels them. It does not ship, operate or configure a log platform, and it does not enforce retention or access rules of its own.
Whatever collects the logs — Loki, Elasticsearch, a SIEM, a cloud logging service — is chosen and configured by whoever runs the observability stack. The labels described below are what makes that configuration possible: they let retention and access be set per kind of log instead of uniformly across everything OLIVE emits.
Log streams and what the labels are for
Logs carry a log_type label separating them into streams, so that each can be retained and access-controlled independently:
| Stream | What it carries |
|---|---|
app | Ordinary application logging |
security | The security event catalogue — authentication and authorisation outcomes, blocked egress, file and job access violations, secret access, configuration changes |
audit_ai | Activity involving AI, marked separately so it can be treated differently from ordinary security logging |
A fourth, developer stream carries the detail developers occasionally need. It is active only in development environments and is never enabled in a production deployment.
The point of the separation is that these three have genuinely different characteristics. security contains personal data — user names, source IP addresses — and is the stream an investigation depends on. audit_ai is the one most likely to be subject to a specific retention obligation. app is ordinary operational noise. Without the label they would share one retention period and one set of access rights, which would be wrong for at least two of them.
What you configure in the log platform
None of this is an OLIVE setting. These are the decisions the labels exist to support, and they belong to whoever configures the observability stack — usually together with the cluster operator.
Retention, per stream. Set it deliberately for each stream rather than accepting one global default. OLIVE does not prescribe a period: how long security logs may or must be kept depends on the compliance regime the operating organisation is subject to, on what the logs contain, and on the data-protection principle that personal data should not be kept longer than necessary. Decide it with the people who own that obligation, and record the reasoning — the number matters less than being able to say why it is that number.
Access, per stream. Restrict the security stream to the people who actually need it. It is not ordinary application logging: it identifies users and their addresses, and broad read access to it is itself a privacy exposure. Everyone who can debug an application should not automatically be able to read who logged in from where. Multi-factor authentication on those accounts is a reasonable additional requirement.
Encryption at rest for log storage. A cluster-operator concern; raise it with them rather than assuming it is covered.
What is and is not logged
This is a rule the services implement, not a setting.
Never logged: credentials, tokens, secrets, API keys, and request or response bodies. A secret's use is recorded — which secret was read or changed, by whom — but never its value. Where developers genuinely need request and response detail, the separate developer stream provides it in development environments only.
Deliberately logged: the security stream records the source IP address of the caller, the user name and the tenant. This is personal data, and it is there on purpose — an event that cannot say who did something from where is of no use for an investigation, and recording it is what several compliance regimes expect of an audit trail.
That is why the security stream is labelled separately: so that it can carry the information an investigation needs, while being retained and access-controlled differently from ordinary application logging. See What you configure in the log platform.
An encrypted-logger approach — encrypting sensitive log lines with a tenant-specific key that operators could decrypt for analysis — was considered and rejected. The lawfulness of logging is determined by what is logged and why, not by how it is stored: encryption reduces the impact of a log breach, but it does not by itself make logging something lawful. The question to answer for each field is whether it is necessary for the purpose — which is why the source IP and user name are in the security stream and request bodies are not. Encryption at rest remains worth having as a protective measure on top of that decision, not as a substitute for it.
The logging mechanism protects itself
A tenant able to reduce logging verbosity globally could suppress the evidence of its own activity. Two controls close that:
- Log level changes no longer affect other tenants' logs.
- Log level and truncation changes are themselves logged, as
OLIMISC1andOLIMISC2.
Verbosity also cannot be raised to a level that captures sensitive data in a production deployment.
What is on every security event
| Field | Notes |
|---|---|
| Event code | For example OLIAUTHN4. Stable across releases — alert on this, not on message text |
| Source IP | Taken one hop behind the proxy for external calls; the calling service name for internal ones. Both protected against forgery |
| User name | |
| Tenant ID | Mandatory metadata on every log record. Without it, logs from different tenants cannot be separated for investigation or for a data subject request |
| Service name and version | Lets an event be attributed to a specific build |
| Description, request ID | A traceable identifier is propagated across service logs, so a single user action can be reconstructed across every service it touched |
Several security events depend on the source IP. If the ingress or gateway in front of OLIVE changes — a different controller, an added load balancer, a different TLS termination point — the propagation and non-spoofability of the client IP must be re-confirmed before those events can be trusted.
Related
- Cluster Operator → Log Events and Automated Reactions — the full event catalogue
- Cluster Operator → Monitoring and Observability