Overview
Wazuh is an open-source security platform. Agents on your Linux, Windows and macOS hosts report to a Wazuh server, which detects intrusions and malware, watches file integrity, finds vulnerable packages and checks configuration against compliance benchmarks. Alerts are stored in the Wazuh indexer and explored in the Wazuh dashboard. This template deploys all three, with agents connecting from anywhere over a public TCP load balancer. Agents are the only ingestion path: the template has no syslog listener. To collect syslog from network devices, run an agent on the host that receives it. For general log search, see the related OpenSearch template.This template deploys into an existing GVC that you already have. It does not create or manage a GVC. Install into a GVC with a single location: the template runs in one location.
What Gets Created
The manager creates a private certificate authority and the TLS certificates on first start. The CA key stays on the manager’s own volume.
Prerequisites
1
Create the credentials secret
One dictionary secret with exactly six keys must exist before you install:
Passwords are 8 to 64 characters, start with a letter, use only
A-Z a-z 0-9 . * + ? -, and contain at least one upper-case letter, lower-case letter, digit and symbol. The Wz1. prefixes above satisfy that. Read the values back with cpln secret reveal SECRET_NAME -o yaml.Set credentialsSecretName to this name.Installation
Install the chart with the credentials secret name:UI
Browse, install, and manage templates visually
CLI
Manage templates from your terminal
Terraform
Declare templates in your Terraform configurations
Pulumi
Declare templates in your Pulumi programs
Configuration
The default is one node of each component. Setindexer.replicas: 3, manager.workers.replicas: 2 and dashboard.replicas: 2 for near-zero-downtime upgrades. Each component also takes an image and a resources block; keep all images on the same Wazuh version.
Credentials
Manager
allowedCidrs to the networks your agents connect from. Edits to ossec.conf made in the dashboard are reset on the next restart; put lasting changes in extraConfig. Custom rules, decoders and CDB lists made in the dashboard are kept. The manager volume holds every agent key and the CA key, so keep its snapshots on.
Workers
Indexer
autoscaling block (enabled, maxCapacity, minFreePercentage, scalingFactor), off by default. Sizing is roughly 2 GiB per agent per 90 days.
Dashboard
Demo Agent
Connecting
Read the credentials with
cpln secret reveal SECRET_NAME -o yaml. The dashboard’s status.canonicalEndpoint and the load balancer host are in each workload’s status:
status.loadBalancer lists tcp://RELEASE_NAME-wazuh-manager-xxxx.yyyy.cpln.app:1514 and …:1515. To open a private dashboard (dashboard.publicAccess.enabled: false), forward it and browse to http://localhost:5601:
Enrolling an Agent
Install the agent on a host with the load balancer host name and the enrollment password, for example on Debian or Ubuntu after adding the Wazuh package repository:Operations
Scaling and Availability
For near-zero-downtime upgrades and failover, setindexer.replicas: 3, manager.workers.replicas: 2 and dashboard.replicas: 2. With the single-node default, a restart of the manager or indexer pauses event processing; agents and the indexer shipper buffer and resend, so events are delayed, not lost. Grow indexer storage with indexer.volumeset.autoscaling. Scale the indexer up, never down: removing nodes can lose the data they hold.
Changing a Password
The secret is the source of truth on every start. Edit the key, then force a redeployment of the indexer, then the manager (and workers), then the dashboard. A password changed only in the dashboard reverts on the next restart.Troubleshooting
Nothing starts and the logs are empty
Nothing starts and the logs are empty
The credentials secret is missing or misnamed. Read the reason from Create the secret or correct
status.versions[].message:credentialsSecretName; the workloads then start on their own.A workload exits with INVALID followed by a key name
A workload exits with INVALID followed by a key name
That key’s password breaks the rules (8 to 64 characters, starts with a letter, only
A-Z a-z 0-9 . * + ? -, one of each kind). Find the key in the logs, fix its value in the secret and force a redeployment of the workload:Vulnerability Detection is empty
Vulnerability Detection is empty
The manager downloads about 1.7 GB of vulnerability content after the first install, which takes two hours or more. If the page is still empty after that, restart the manager once (Server management, Status, Restart).
An agent shows as disconnected after the demo agent restarts
An agent shows as disconnected after the demo agent restarts
The demo agent enrolls under a new name each time it restarts and can leave the old entry disconnected. Remove stale entries under Endpoints, and turn the demo agent off once your own agents connect.
Important Notes
- Agent ports are open to the whole GVC while the load balancer is on.
manager.internalAccessdoes not restrict ports1514and1515whenmanager.publicAccess.enabledistrue; the enrollment password is the gate. It restricts them (and the API) once the load balancer is off. manager.internalAccess.typecannot benone: the dashboard needs the API.- Custom rules in
extraConfig. A<ruleset>block replaces the default one, so restate the defaultdecoder_dir,rule_dirandlistentries in it, or the default rules stop loading. Workers load rules synced from the master after a restart (PUT /cluster/restart). - Restarts can repeat alerts. Agents resend events the manager had not acknowledged, so an alert can appear twice after a restart. The Statistics page returns errors for about an hour after a manager start.
- Access changes take time. After changing
manager.publicAccess.*ordashboard.publicAccess.enabled, allow 30 seconds to about 10 minutes for the firewall and load balancer to apply. - Never replace the manager volume set. It holds every agent key and the CA key. If it is ever replaced, restart the indexer and dashboard so they pick up the new CA, and enroll the agents again.
- Uninstalling deletes every volume set, with the stored alerts and every agent key; a final snapshot of each is kept for
retentionDuration. - Disk. The indexer’s disk watermarks are off (as upstream ships it), so a full volume stops indexing. Keep the retention values set or turn on volume autoscaling.
- Indexer node count must be
1or an odd number of3or more.