Defaults work for most installs. This page lists what you can change.
Helm chart watchlog (server and agents)
| Value | Default | What it does |
|---|---|---|
publicUrl |
'' |
The URL people use to open WatchLog, used in alert links and install commands. |
admin.email, admin.password |
'' |
Create the admin at install time instead of using the setup code. |
secretKey |
generated | Key for alert channel secrets (32 bytes, base64). Set it only to restore a backup. |
clusterName |
kubernetes |
Name of this cluster in WatchLog. |
image.registry, image.tag |
ghcr.io/watchlog-light, chart version |
Where WatchLog images come from: use a mirror for air-gapped installs. |
backend.replicas, backend.resources |
1, 192Mi–512Mi | API server. |
backend.eventsRetentionDays |
15 | How long Kubernetes events are kept. |
alertWorker.replicas |
1 | Evaluates rules and sends notifications; replicas share the work. |
alerting.smtpUrl, alerting.smtpFrom |
'' |
SMTP server for email alerts, e.g. smtp://user:[email protected]:587. |
alerting.telegramApiUrl |
Telegram's | Telegram Bot API mirror, where api.telegram.org is blocked. |
alerting.httpsProxy, alerting.noProxy |
'' |
Outbound proxy for Slack, Telegram and webhook alerts. |
web.service.type, web.service.port |
ClusterIP, 8080 |
The UI's Service. |
ingress.enabled, ingress.host, ingress.className, ingress.tlsSecret, ingress.annotations |
off | Expose the UI through an Ingress. |
postgresql.storage, postgresql.storageClass |
8Gi | Database volume. |
victoriametrics.retention, victoriametrics.storage |
15d, 20Gi |
Metrics retention and volume. |
watchlog-agent.enabled |
true |
Install the agents for this cluster. |
watchlog-agent.* |
see below | Agent settings, prefixed with watchlog-agent.. |
Helm chart watchlog-agent (agents only)
| Value | Default | What it does |
|---|---|---|
watchlog.endpoint |
required | The WatchLog URL the agents push to. |
watchlog.clusterToken or watchlog.existingSecret |
required | The cluster token from Settings → Clusters, or a Secret that holds it (existingSecretKey, default token). |
agent.networkFlows.enabled |
true |
eBPF network flows for the service map. When on, the node agent runs as uid 0 with only the BPF and PERFMON capabilities; when off, as uid 65532 with none. See security. |
agent.interval |
15s |
Collection interval. |
agent.kubeletTLS |
auto |
Kubelet certificate check: auto verifies against the cluster CA and falls back if that fails, verify never falls back, skip never verifies. |
agent.resources |
48Mi request, 128Mi limit | Node agent resources. It stays within 64 MB of memory and 1 % of a core. |
agent.tolerations, agent.nodeSelector, agent.priorityClassName |
every node | Where the node agent runs. |
clusterAgent.enabled, clusterAgent.resources, clusterAgent.goMemLimit |
on, 128Mi–512Mi, 400MiB | The cluster-agent that watches the Kubernetes API. |
image.registry, image.tag, image.pullPolicy |
ghcr.io/watchlog-light, chart version |
Agent images. |
Docker Compose server (.env)
| Variable | Default | What it does |
|---|---|---|
WATCHLOG_DB_PASSWORD |
required | PostgreSQL password. |
WATCHLOG_SECRET_KEY |
required | Key for alert channel secrets: openssl rand -base64 32. Back it up. |
WATCHLOG_PUBLIC_URL |
'' |
How agents and browsers reach the server. |
WATCHLOG_VERSION |
latest |
Image version; pin it, e.g. 0.1.0. |
WATCHLOG_PORT |
8080 | The only published port. |
WATCHLOG_RETENTION |
15d |
Metrics retention. |
WATCHLOG_SMTP_URL, WATCHLOG_SMTP_FROM |
unset | Email alerts. |
HTTPS_PROXY, NO_PROXY |
unset | Outbound proxy for alert deliveries. |
WATCHLOG_TELEGRAM_API_URL |
Telegram's | Telegram Bot API mirror. |
WATCHLOG_ADMIN_EMAIL, WATCHLOG_ADMIN_PASSWORD |
unset | Create the admin at startup instead of the setup code. |
Agent options
Set as environment variables, or as flags in the docker run command. The install commands set the required ones.
| Variable | Flag | Default | What it does |
|---|---|---|---|
WATCHLOG_ENDPOINT |
--endpoint |
required | The WatchLog URL. |
WATCHLOG_TOKEN |
— | required | The cluster token. |
WATCHLOG_EBPF |
--ebpf |
true |
Network flows. false turns them off; the service map then says so for this agent. |
WATCHLOG_HEALTH_OUTPUT |
--health-output |
true |
Docker: send health check output (redacted, one line, at most 512 bytes). false sends exit codes only. |
WATCHLOG_HOSTNAME |
--hostname |
the host's | The name the host appears under. |
WATCHLOG_DOCKER_SOCKET |
--docker-socket |
/var/run/docker.sock |
Docker socket, used read-only. |
WATCHLOG_BUFFER_DIR |
--buffer-dir |
/var/lib/watchlog-agent/buffer |
On-disk buffer while the server is unreachable: up to 200 MB or 15 minutes. |