Wazuh consulting · Guide · Graylog
Wazuh and Graylog:running Wazuh at scale.
Our own SOC keeps the Wazuh manager as its signal and alert engine throughout, but stores the data in Graylog in place of the Wazuh indexer, and reads it through a front end we wrote ourselves. This guide explains why we built it that way, how the parts fit together, the problems the community forums are full of, what you give up, and when the standard Wazuh design is the better choice.
Wazuh 4.14Graylog 7Updated November 2026From a design in production
Keep the Wazuh manager for detection. Put Graylog where the indexer was when the data, the retention or the number of tenants outgrows it.
The Wazuh manager is where detection happens: agents and collectors feed it, and its decoders and rules decide what becomes an alert. The indexer and dashboard are storage and display, and they can be replaced. Graylog adds a mature ingestion layer, streams, pipelines and index management in front of OpenSearch, which is what a SOC serving many clients needs. For a single organisation with ordinary volumes, the standard Wazuh design is simpler and should be the default.
- Detection
- Wazuh manager, rules and decoders unchanged
- Storage
- Graylog, with OpenSearch managed by its Data Node
- Display
- Your own front end, or Graylog's search and dashboards
- Suits
- Many tenants, high volumes, long retention
Why
Why replace the Wazuh indexer with Graylog?
The Wazuh indexer is OpenSearch packaged for Wazuh, and for most deployments it is the right store. We moved away from it for reasons that come with running a multi-tenant SOC across eleven countries:
Ingestion as a product, not a pipe. Graylog receives data through inputs that buffer to a journal on disk, so a slow or unavailable storage layer does not lose events. That journal is also an early warning: when it grows, something downstream is wrong.
Streams and index sets. Each client's data can be routed by streams into its own index sets, with its own rotation and retention. Tenancy is then a matter of which indices you search, which is simpler to reason about and to prove than a filter on a field.
Pipelines. Graylog's processing pipelines enrich, normalise and route events before they are stored: adding asset and client context, dropping noise, and sending a copy elsewhere when a client asks for it.
One store for everything. Wazuh alerts sit beside firewall logs, cloud audit data and anything else that arrives by syslog or API, so analysts and automation read from one place.
Retention at scale. Index sets rotate and age on rules you set, and OpenSearch nodes can be arranged in hot and warm tiers so recent data is fast and older data is cheaper to keep.
Graylog's free edition, Graylog Open, is free to self-host but is licensed under the Server Side Public License rather than an open source licence approved by the Open Source Initiative, which is worth knowing before you build on it. Some features, such as its data lake archive search, are in the paid editions.
Architecture
How the parts fit together
| Layer | Standard Wazuh | Wazuh with Graylog |
|---|---|---|
| Collection | Wazuh agents, syslog, cloud modules | The same |
| Detection | Wazuh manager: decoders and rules | The same, unchanged |
| Shipping | Filebeat sends alerts to the Wazuh indexer | A shipper or syslog output sends alerts to a Graylog input |
| Processing | None between manager and store | Graylog streams and pipelines |
| Storage | Wazuh indexer (OpenSearch) | OpenSearch managed by Graylog's Data Node |
| Display | Wazuh dashboard | Graylog, or your own front end |
| State data | Written by the manager to the indexer | Needs a decision: see below |
Graylog's Data Node manages OpenSearch for Graylog: it issues certificates, handles cluster membership and keeps the OpenSearch version compatible. Graylog 7 is the current major release; we keep the Data Node and the Graylog servers on the same release, upgraded together.
Shipping alerts
Getting Wazuh alerts into Graylog
The manager writes every alert as a line of JSON to /var/ossec/logs/alerts/alerts.json. There are two sensible ways to move them.
A log shipper reading alerts.json. Fluent Bit, Filebeat or Graylog's own Sidecar tails the file and sends each line to a Graylog input (Beats, GELF or raw over TCP with TLS). The shipper remembers its position, so a restart or a network break resumes where it stopped. This is the method we recommend.
Wazuh's syslog output. The manager can send alerts itself in JSON to a syslog input, filtered by level, group or rule:
<syslog_output>
<server>graylog.example.internal</server>
<port>5514</port>
<level>3</level>
<format>json</format>
</syslog_output>
It is the simplest to set up and needs nothing on the manager besides the configuration, but it is less forgiving of interruptions than a shipper that tracks its place in the file. Use it for a modest volume, or to send a filtered copy of alerts somewhere else.
Alerts or everything? alerts.json holds only events that matched a rule. If you want every event the manager receives, for hunting or for evidence, turn on <logall_json>yes</logall_json> in the global section and ship /var/ossec/logs/archives/archives.json too. It is far larger; give it its own index set and its own, usually shorter, retention.
In a cluster, ship from every manager node, not only the master: each worker writes the alerts for the agents connected to it.
Field names
The field name problem, and how to handle it
Wazuh's alerts use nested fields written with dots: rule.level, agent.id, data.win.eventdata.targetUserName. In our design the Graylog pipeline flattens them so that they are stored with underscores instead: rule_level, agent_id, agent_name. Graylog users on its community forum report the same behaviour and the same consequence: the Wazuh dashboard, which expects Wazuh's own field names and index patterns, no longer shows the data. That is the moment people find they need another way to display alerts.
Plan for it from the start:
- Decide the names once. Write down how every Wazuh field you rely on is stored, and keep that list with the pipeline rules.
- Rewrite searches and dashboards. Anything that queried
rule.levelnow queriesrule_level. Saved searches, alert definitions, reports and integrations all need checking. - Keep the raw alert. Store the original JSON in a field as well, so an investigator can always see exactly what Wazuh produced.
- Think ahead to 5.0. Wazuh 5.0 moves to a new schema based on the Elastic Common Schema, which will change the field names arriving at your pipelines again. A readiness review covers the pipelines as well as the rules.
What you lose
What you give up, and the decision about state data
Replacing the indexer is not free. Be clear about these before you start:
The Wazuh dashboard's modules. The views for agents, configuration assessment, file integrity, vulnerabilities and compliance read Wazuh's own indices and API. With the data in Graylog, you rebuild the views you need there or in a front end of your own, as we did.
State data. Since Wazuh 4.8 the manager writes some data directly to the indexer through the <indexer> block in ossec.conf, most visibly the vulnerability detection results. A Graylog design has to decide where that goes: a small Wazuh indexer kept for state data, or a deliberate choice to read vulnerability and inventory data through the Wazuh API instead. Leaving the <indexer> block pointing at nothing, or at the wrong address, produces errors in ossec.log and missing vulnerability data.
Documentation and community help. Most Wazuh guidance assumes the standard design. When something goes wrong, fewer people have seen your exact setup.
More to run. Graylog servers, MongoDB for Graylog's configuration, the Data Node and OpenSearch, each with its own upgrades.
From the engine room
Problems we have met running Wazuh with Graylog
- _01A journal that would not drainA DNS fault poisoned a connection pool between Graylog and its storage, and twenty million messages queued in the journal. Nothing was lost because the journal held them; the fix was in name resolution, and journal size is now watched.Ingestion
- _02Replica shards that never assignedAn index lifecycle policy moved older indices to warm nodes but left their replicas permanently unassigned, so the cluster sat at yellow. Tiering rules and a nightly enforcement job fixed it.Storage
- _03The wrong address in the indexer blockA manager's `<indexer>` block pointed at a wildcard bind address rather than a real host, so state data never arrived. A one-line fix, found only by reading `ossec.log`.State data
- _04Enrolment storms through the load balancerEnrolment traffic on port 1515 went to worker nodes instead of the master, and thousands of agents retried at once. Enrolment now goes only to the node that issues keys.Clusters
- _05Fields that changed name under a dashboardSearches written against dotted Wazuh field names returned nothing once the pipeline flattened them. The fix was a field map, agreed once and used everywhere.Fields
When not to
When the standard Wazuh design is the better choice
For most single organisations, keep Wazuh's own indexer and dashboard. You get the full dashboard, the documentation matches your deployment, and there is one less platform to run. Our installation guide covers that design step by step, and our deployment service builds it at a fixed price.
Consider Graylog when one or more of these is true: you serve several tenants and need their data kept apart; you need to keep a year or more of data and the indexer's storage is becoming the cost; you already run Graylog for other logs; or you need a processing layer between detection and storage. If you are not sure, a health check of your current deployment will tell you whether it has outgrown the standard design.
If you are still choosing a SIEM at all, our comparison of Wazuh with Splunk, Sentinel and Elastic may help first.
How we help
Wazuh and Graylog work we do
Design
Whether Graylog is worth it for you, and if so the layout: inputs, streams, index sets, retention, tiers and the state-data decision.
Build or migrate
A new Wazuh and Graylog deployment, or a move from the Wazuh indexer with old and new run side by side. See deployment.
Pipelines and field maps
Pipeline rules that normalise Wazuh's fields, add context and route each tenant's data, with the field map documented.
Support
A support plan that covers the Graylog layer we built as well as Wazuh, from business hours to 24/7 for critical issues.
When it breaks
Journals that will not drain, shards that will not assign, alerts that stopped arriving. Start with our troubleshooting guide.
Questions
Questions about Wazuh and Graylog
Can Graylog replace the Wazuh indexer?
Yes. The Wazuh manager does the detection; the indexer stores the results. Alerts can be shipped from the manager to Graylog instead, which stores them in OpenSearch through its Data Node. You lose the Wazuh dashboard's own views and need to decide where the manager's state data goes.
How do I send Wazuh alerts to Graylog?
Either run a log shipper such as Fluent Bit, Filebeat or Graylog Sidecar that reads /var/ossec/logs/alerts/alerts.json and sends it to a Graylog input, or configure the manager's syslog output in JSON format. A shipper is more resilient; syslog output is simpler.
Why does the Wazuh dashboard stop working with Graylog?
Because the dashboard expects Wazuh's own indices and field names. Graylog stores fields with underscores rather than dots and in its own index sets, so the dashboard's views find nothing. You display the data in Graylog or a front end of your own.
Is Graylog free?
Graylog Open is free to self-host, under the Server Side Public License. Graylog's paid editions add further features, such as its data lake archive search.
Does your SOC use Wazuh?
Yes. The Wazuh manager is the signal and alert engine throughout our SOC; Graylog replaces the Wazuh indexer, for better scaling at the data lake level; and the front end is our own.
Should we move to Graylog before Wazuh 5.0?
Plan them together. Wazuh 5.0 changes the alert schema, so pipelines built now will need revisiting. A readiness review covers both.
Wazuh and Graylog
Thinking about Graylog for your Wazuh?
Thirty minutes on your volumes, tenants and retention, and whether it is worth it. We will tell you if it is not.