MariaDB¶

Connect HolmesGPT to MariaDB databases to analyze query performance, investigate slow queries, check replication status, examine database health, and read data for troubleshooting.

You can configure multiple MariaDB instances with different names (e.g., app-mariadb, cache-mariadb, staging-mariadb).

Creating a Read-Only User¶

-- Create user
CREATE USER 'holmes_readonly'@'%' IDENTIFIED BY 'your_secure_password';

-- Grant read-only permissions
GRANT SELECT, SHOW VIEW, PROCESS, REPLICATION CLIENT ON *.* TO 'holmes_readonly'@'%';

-- Grant access to performance and information schemas
GRANT SELECT ON performance_schema.* TO 'holmes_readonly'@'%';
GRANT SELECT ON information_schema.* TO 'holmes_readonly'@'%';

FLUSH PRIVILEGES;

Configuration¶

Connection URL format:

mysql+pymysql://[username]:[password]@[host]:[port]/[database]

Note: MariaDB uses MySQL wire protocol, so use mysql+pymysql:// in the connection URL.

Holmes CLIHolmes Helm ChartRobusta Helm Chart

~/.holmes/config.yaml:

toolsets:
  app-mariadb:
    type: database
    config:
      connection_url: "mysql+pymysql://holmes_readonly:your_secure_password@mariadb.example.com:3306/appdb"
    llm_instructions: "Application database with user and session data"

  cache-mariadb:
    type: database
    config:
      connection_url: "mysql+pymysql://cache_user:pass@cache-mariadb.internal:3306/cache"
    llm_instructions: "Cache database for session storage"

Using environment variables:

toolsets:
  app-mariadb:
    type: database
    config:
      connection_url: "{{ env.MARIADB_URL }}"

Create a Kubernetes secret in the namespace Holmes runs in:

kubectl create secret generic holmes-database-mariadb \
  --from-literal=MARIADB_URL='mysql+pymysql://holmes_readonly:your_secure_password@mariadb.example.com:3306/appdb' \
  -n <namespace>

When using the standalone Holmes Helm Chart, update your values.yaml:

extraEnvVarsSecrets:
  - holmes-database-mariadb

toolsets:
  app-mariadb:
    type: database
    config:
      connection_url: "{{ env.MARIADB_URL }}"
    llm_instructions: "Application database with user and session data"

Apply the configuration:

helm upgrade holmes robusta/holmes -f values.yaml

Create a Kubernetes secret in the namespace Holmes runs in:

kubectl create secret generic holmes-database-mariadb \
  --from-literal=MARIADB_URL='mysql+pymysql://holmes_readonly:your_secure_password@mariadb.example.com:3306/appdb' \
  -n <namespace>

When using the Robusta Helm Chart (which includes HolmesGPT), update your generated_values.yaml:

holmes:
  extraEnvVarsSecrets:
    - holmes-database-mariadb

  toolsets:
    app-mariadb:
      type: database
      config:
        connection_url: "{{ env.MARIADB_URL }}"
      llm_instructions: "Application database with user and session data"

Apply the configuration:

helm upgrade robusta robusta/robusta -f generated_values.yaml --set clusterName=<YOUR_CLUSTER_NAME>

Multiple instances¶

Holmes Helm ChartRobusta Helm Chart

Create a Kubernetes secret in the namespace Holmes runs in:

kubectl create secret generic holmes-database-mariadb-instances \
  --from-literal=APP_MARIADB_URL='mysql+pymysql://holmes_readonly:your_secure_password@mariadb.example.com:3306/appdb' \
  --from-literal=CACHE_MARIADB_URL='mysql+pymysql://cache_user:pass@cache-mariadb.internal:3306/cache' \
  -n <namespace>

When using the standalone Holmes Helm Chart, update your values.yaml:

extraEnvVarsSecrets:
  - holmes-database-mariadb-instances

toolsets:
  app-mariadb:
    type: database
    config:
      connection_url: "{{ env.APP_MARIADB_URL }}"

  cache-mariadb:
    type: database
    config:
      connection_url: "{{ env.CACHE_MARIADB_URL }}"

Apply the configuration:

helm upgrade holmes robusta/holmes -f values.yaml

Create a Kubernetes secret in the namespace Holmes runs in:

kubectl create secret generic holmes-database-mariadb-instances \
  --from-literal=APP_MARIADB_URL='mysql+pymysql://holmes_readonly:your_secure_password@mariadb.example.com:3306/appdb' \
  --from-literal=CACHE_MARIADB_URL='mysql+pymysql://cache_user:pass@cache-mariadb.internal:3306/cache' \
  -n <namespace>

When using the Robusta Helm Chart (which includes HolmesGPT), update your generated_values.yaml:

holmes:
  extraEnvVarsSecrets:
    - holmes-database-mariadb-instances

  toolsets:
    app-mariadb:
      type: database
      config:
        connection_url: "{{ env.APP_MARIADB_URL }}"

    cache-mariadb:
      type: database
      config:
        connection_url: "{{ env.CACHE_MARIADB_URL }}"

Apply the configuration:

helm upgrade robusta robusta/robusta -f generated_values.yaml --set clusterName=<YOUR_CLUSTER_NAME>

Configuration Options¶

connection_url (required): MariaDB connection URL
read_only (default: true): Only allow SELECT/SHOW/DESCRIBE/EXPLAIN/WITH statements
verify_ssl (default: true): Verify SSL certificates
max_rows (default: 200): Maximum rows to return (1-10000)
llm_instructions: Context about this database

Common Use Cases¶

"Analyze query performance: SELECT * FROM users WHERE last_login > NOW() - INTERVAL 30 DAY"

"Show replication status and lag"

"List tables by size"

Migrating from the MariaDB MCP addon¶

Earlier versions shipped a separate MariaDB MCP server, enabled through mcpAddons.mariadb in the Helm chart. That addon has been removed and the chart no longer renders it, so leaving the old value in place gives you no MariaDB access at all.

To migrate, remove the mcpAddons.mariadb block from your Helm values and configure this data source with your connection URL as shown above. The MCP server's read-only mode is replaced by the read_only option, which defaults to true.