Docker launcher

The virtuoso-launch command provides a convenient way to launch a Virtuoso database using Docker with various configurable parameters.

Basic usage

# With pipx (global installation)
virtuoso-launch

# With uv (development)
uv run python virtuoso_utilities/launch_virtuoso.py

This launches a Virtuoso container with default settings.

Customized usage

virtuoso-launch \
    --name my-virtuoso \
    --http-port 8891 \
    --isql-port 1112 \
    --data-dir ./my-virtuoso-data \
    --dba-password mySafePassword \
    --mount-volume /path/on/host/with/rdf:/rdf-data-in-container \
    --network my-docker-network \
    --memory 16g \
    --detach \
    --wait-ready \
    --enable-write-permissions

Arguments

Use virtuoso-launch --help to see all available options:

Argument	Description	Default
--name	Name for the Docker container	virtuoso
--http-port	HTTP port to expose Virtuoso on	8890
--isql-port	ISQL port to expose Virtuoso on	1111
--data-dir	Host directory to mount as Virtuoso data directory	./virtuoso-data
--mount-volume	Mount additional host directory (format: HOST:CONTAINER). Can be specified multiple times	-
--memory	Memory limit for the container (e.g., 2g, 4g)	~2/3 of host RAM
--cpu-limit	CPU limit for the container (0 = no limit)	0
--dba-password	Password for the Virtuoso dba user	dba
--max-dirty-buffers	Maximum dirty buffers before checkpoint	Auto-calculated
--number-of-buffers	Number of buffers	Auto-calculated
--estimated-db-size-gb	Estimated database size in GB for preconfiguring MaxCheckpointRemap	-
--network	Docker network to connect the container to	-
--wait-ready	Wait until Virtuoso is ready to accept connections	false
--enable-write-permissions	Enable write permissions for ‘nobody’ and ‘SPARQL’ users	false
--detach	Run container in detached mode	false
--force-remove	Force removal of existing container with the same name	false
--virtuoso-version	Virtuoso Docker image version/tag	latest
--virtuoso-sha	Virtuoso Docker image SHA256 digest (takes precedence over version)	-
--parallel-threads	Maximum parallel threads for query execution. If not specified, uses all available CPU cores	Auto-detected

Memory-based configuration

Docker resource limits

By default, Docker Desktop runs with limited resources. Before allocating significant memory using the --memory flag, ensure Docker is configured to have access to at least that amount of RAM. If the Virtuoso container requests more memory than Docker can provide, it may fail to start or crash (often with exit code 137).

The script simplifies memory configuration based on Virtuoso best practices:

Container memory limit

The script automatically detects your host’s total RAM and sets the default --memory to approximately 2/3 of that total. This follows the Virtuoso guideline of allocating a significant portion of system RAM to the database process.

Docker memory reservation

To prevent out-of-memory (OOM) crashes, the script sets Docker’s --memory-reservation to 85% of the --memory limit, providing a 15% safety margin.

Virtuoso internal buffers

The script calculates optimal values for NumberOfBuffers and MaxDirtyBuffers based on 85% of the container memory limit:

EffectiveMemory = ContainerMemoryLimit * 0.85
NumberOfBuffers = (EffectiveMemory * 0.66) / 8700
MaxDirtyBuffers = NumberOfBuffers * 0.75

These formulas are derived from official OpenLink documentation:

• Buffer size (8700 bytes): According to Performance diagnostics: “Each buffer caches one 8K page of data and occupies approximately 8700 bytes of memory.”

• Memory allocation (66%): According to RDF Performance Tuning: “Adding more than 60-70% of system ram as buffers is not useful.”

• MaxDirtyBuffers (75%): According to RDF Performance Tuning: “Typical sizes for the NumberOfBuffers and MaxDirtyBuffers (3/4 of NumberOfBuffers) parameters”

Example with --memory 10g:

• Docker hard limit: 10g (100%)
• Docker soft limit (reservation): 8.5g (85%)
• Virtuoso buffer calculations: based on 8.5g
• Available headroom: 1.5g (15%) for process overhead

Allowed directories

The script automatically constructs the VIRT_Parameters_DirsAllowed environment variable, including the container data directory and any paths specified via --mount-volume.

Automatic MaxCheckpointRemap configuration

For large databases (> 1 GiB), Virtuoso recommends tuning the MaxCheckpointRemap parameter. This script offers two methods:

For existing databases (default)

1. The script checks for an existing virtuoso.ini file in --data-dir
2. If the total size exceeds 1 GiB, it calculates the recommended value (1/4th of total size in 8K pages)
3. It modifies MaxCheckpointRemap in both [Database] and [TempDatabase] sections

For new deployments

Use --estimated-db-size-gb to preconfigure MaxCheckpointRemap via environment variables:

virtuoso-launch --estimated-db-size-gb 100

Query parallelization

The script automatically configures Virtuoso for parallel query execution based on available CPU cores.

Threading parameters

When launched, the script sets:

AsyncQueueMaxThreads = CPU_cores * 1.5
ThreadsPerQuery = CPU_cores
MaxClientConnections = CPU_cores * 2
HTTPServer_ServerThreads = CPU_cores * 2

You can override the detected CPU count using --parallel-threads:

virtuoso-launch --parallel-threads 8

Query memory

The script calculates MaxQueryMem to prevent out-of-memory errors during query execution:

MaxQueryMem = (EffectiveMemory - BufferMemory) * 0.8

This ensures queries have dedicated memory separate from the buffer pool.

Adaptive vector sizing

The script enables adaptive vector sizing for better performance on large queries:

AdjustVectorSize = 1
MaxVectorSize = 1000000

Documentation sources

• AsyncQueueMaxThreads: According to Virtuoso configuration tips: “This should be set to either 1.5 * the number of cores or 1.5 * the number of core threads; see which works better.”

• ThreadsPerQuery: According to Configuration Parameters for Vectoring and Parallelization: “The number of cores on the machine is a reasonable default if running large queries.”

• MaxClientConnections: According to OpenLink Community: “sets the maximum number of threads that can be allocated for SQL Connectivity”. Set to CPU_cores * 2 to ensure it exceeds AsyncQueueMaxThreads.

• AdjustVectorSize: According to Configuration Parameters for Vectoring and Parallelization: “Set AdjustVectorSize = 1 to enable this feature. The SQL execution engine will increase the vector size when it sees an index lookup that does not get good locality.” This can yield “up to a 3x speed-up” on large queries.

• MaxVectorSize: According to Configuration Parameters for Vectoring and Parallelization: “When AdjustVectorSize is on, this setting gives the maximum vector size. The default is 1,000,000 and the largest allowed value is about 3,500,000.”

Virtuoso 7.0+

These parallelization settings only apply to Virtuoso 7.0 and later versions.

Client SQL timeouts

To ensure reliable operation of long-running operations, the launcher enforces these settings in [Client] of virtuoso.ini:

• SQL_QUERY_TIMEOUT = 0
• SQL_TXN_TIMEOUT = 0

Programmatic usage

The launch_virtuoso function can be imported and called directly from Python code:

from virtuoso_utilities.launch_virtuoso import launch_virtuoso

launch_virtuoso(
    name="my-virtuoso",
    data_dir="./virtuoso-data",
    memory="8g",
    dba_password="dba",
    detach=True,
    wait_ready=True,
    enable_write_permissions=True,
)

The function parameters correspond to the CLI arguments described above.