Presto C++ Configuration Properties¶

This section describes Presto C++ configuration properties.

The following is not a complete list of all configuration properties, and does not include any connector-specific catalog configuration properties or session properties.

For information on catalog configuration properties, see Connectors.

For information on Presto C++ session properties, see Presto C++ Session Properties.

NOTE: While some of the configuration properties below with “-gb” in their names show gigabytes (gB; 1 gB equals 1000000000 B), it is actually gibibytes (GiB; 1 GiB equals 1073741824 B).

Coordinator Properties¶

Set the following configuration properties for the Presto coordinator exactly as they are shown in this code block to enable the Presto coordinator’s use of Presto C++ workers.

native-execution-enabled=true
optimizer.optimize-hash-generation=false
regex-library=RE2J
use-alternative-function-signatures=true

These Presto coordinator configuration properties are described here, in alphabetical order.

driver.cancel-tasks-with-stuck-operators-threshold-ms¶

• Type: string

• Default value: 240000 (40 minutes)

  Cancels any task when at least one operator has been stuck for at least the time specified by this threshold.

  Set this property to 0 to disable canceling.

native-execution-enabled¶

• Type: boolean

• Default value: false

  This property is required when running Presto C++ workers because of underlying differences in behavior from Java workers.

optimizer.optimize-hash-generation¶

• Type: boolean

• Default value: true

  Set this property to false when running Presto C++ workers. Velox does not support optimized hash generation, instead using a HashTable with adaptive runtime optimizations that does not use extra hash fields.

regex-library¶

• Type: type

• Allowed values: RE2J

• Default value: JONI

  Only RE2J is currently supported by Velox.

use-alternative-function-signatures¶

• Type: boolean

• Default value: false

  Some aggregation functions use generic intermediate types which are not compatible with Velox aggregation function intermediate types. One example function is approx_distinct, whose intermediate type is VARBINARY. This property provides function signatures for built-in aggregation functions which are compatible with Velox.

presto.default-namespace¶

• Type: string

• Default value: presto.default

  Specifies the namespace prefix for native C++ functions.

Worker Properties¶

The configuration properties of Presto C++ workers are described here, in alphabetical order.

runtime-metrics-collection-enabled¶

• Type: boolean

• Default value: false

  Enables collection of worker level metrics.

task.max-drivers-per-task¶

• Type: integer

• Default value: number of concurrent threads supported by the host

  Number of drivers to use per task. Defaults to the number of concurrent threads supported by the host.

query.max-memory-per-node¶

• Type: string

• Default value: 4GB

  Max memory usage for each query.

system-memory-gb¶

• Type: integer

• Default value: 57

  Memory allocation limit enforced by an internal memory allocator. It consists of two parts: 1) Memory used by the queries as specified in query-memory-gb; 2) Memory used by the system, such as disk spilling and cache prefetch.

  Set system-memory-gb to about 90% of available machine memory of the deployment. This allows some buffer room to handle unaccounted memory in order to prevent out-of-memory conditions. The default value of 57 gb is calculated based on available machine memory of 64 gb.

query-memory-gb¶

• Type: integer

• Default value: 38

  Specifies the total amount of memory in GB that can be used for all queries on a worker node. Memory for system usage such as disk spilling and cache prefetch are not counted in it.

max_spill_bytes¶

• Type: integer

• Default value: 100UL << 30

  Specifies the max spill bytes limit set for each query. This is used to cap the storage used for spilling. If it is zero, then there is no limit and spilling might exhaust the storage or takes too long to run.

spill-enabled¶

• Type: boolean

• Default value: false

Try spilling memory to disk to avoid exceeding memory limits for the query.

Spilling works by offloading memory to disk. This process can allow a query with a large memory footprint to pass at the cost of slower execution times. Currently, spilling is supported only for aggregations and joins (inner and outer), so this property will not reduce memory usage required for window functions, sorting and other join types.

join-spill-enabled¶

• Type: boolean

• Default value: true

When spill_enabled is true, this determines whether Presto will try spilling memory to disk for joins to avoid exceeding memory limits for the query.

aggregation-spill-enabled¶

• Type: boolean

• Default value: true

When spill_enabled is true, this determines whether Presto will try spilling memory to disk for aggregations to avoid exceeding memory limits for the query.

order-by-spill-enabled¶

• Type: boolean

• Default value: true

When spill_enabled is true, this determines whether Presto will try spilling memory to disk for order by to avoid exceeding memory limits for the query.

shared-arbitrator.reserved-capacity¶

• Type: string

• Default value: 4GB

  Specifies the total amount of memory in GB reserved for the queries on a worker node. A query can only allocate from this reserved space if 1) the non-reserved space in query-memory-gb is used up; and 2) the amount it tries to get is less than shared-arbitrator.memory-pool-reserved-capacity.

shared-arbitrator.memory-pool-initial-capacity¶

• Type: string

• Default value: 128MB

  The initial memory pool capacity in bytes allocated on creation.

shared-arbitrator.global-arbitration-enabled¶

• Type: string

• Default value: false

  If true, it allows shared arbitrator to reclaim used memory across query memory pools.

shared-arbitrator.memory-pool-reserved-capacity¶

• Type: string

• Default value: 64MB

  The amount of memory in bytes reserved for each query memory pool. When a query tries to allocate memory from the reserved space whose size is specified by shared-arbitrator.reserved-capacity, it cannot allocate more than the value specified in shared-arbitrator.memory-pool-reserved-capacity.

shared-arbitrator.memory-pool-transfer-capacity¶

• Type: string

• Default value: 32MB

  The minimal memory capacity in bytes transferred between memory pools during memory arbitration.

shared-arbitrator.max-memory-arbitration-time¶

• Type: string

• Default value: 5m

  Specifies the max time to wait for memory reclaim by arbitration. The memory reclaim might fail if the max wait time has exceeded. If it is zero, then there is no timeout.

shared-arbitrator.fast-exponential-growth-capacity-limit¶

• Type: string

• Default value: 512MB

  When shared arbitrator grows memory pool’s capacity, the growth bytes will be adjusted in the following way:

  • If 2 * current capacity is less than or equal to shared-arbitrator.fast-exponential-growth-capacity-limit, grow through fast path by at least doubling the current capacity, when conditions allow (see below NOTE section).

  • If 2 * current capacity is greater than shared-arbitrator.fast-exponential-growth-capacity-limit, grow through slow path by growing capacity by at least shared-arbitrator.slow-capacity-grow-pct * current capacity if allowed (see below NOTE section).

  NOTE: If original requested growth bytes is larger than the adjusted growth bytes or adjusted growth bytes reaches max capacity limit, the adjusted growth bytes will not be respected.

  NOTE: Capacity growth adjust is only enabled if both shared-arbitrator.fast-exponential-growth-capacity-limit and shared-arbitrator.slow-capacity-grow-pct are set, otherwise it is disabled.

shared-arbitrator.slow-capacity-grow-pct¶

• Type: string

• Default value: 0.25

  See description for shared-arbitrator.fast-exponential-growth-capacity-limit

shared-arbitrator.memory-pool-min-free-capacity¶

• Type: string

• Default value: 128MB

  When shared arbitrator shrinks memory pool’s capacity, the shrink bytes will be adjusted in a way such that AFTER shrink, the stricter (whichever is smaller) of the following conditions is met, in order to better fit the pool’s current memory usage:

  • Free capacity is greater or equal to capacity * shared-arbitrator.memory-pool-min-free-capacity-pct

  • Free capacity is greater or equal to shared-arbitrator.memory-pool-min-free-capacity

  NOTE: In the conditions when original requested shrink bytes ends up with more free capacity than above two conditions, the adjusted shrink bytes is not respected.

  NOTE: Capacity shrink adjustment is enabled when both shared-arbitrator.memory-pool-min-free-capacity-pct and shared-arbitrator.memory-pool-min-free-capacity are set.

shared-arbitrator.memory-pool-min-free-capacity-pct¶

• Type: string

• Default value: 0.25

  See description for shared-arbitrator.memory-pool-min-free-capacity

shared-arbitrator.memory-pool-abort-capacity-limit¶

• Type: string

• Default value: 1GB

  Specifies the starting memory capacity limit for global arbitration to search for victim participant to reclaim used memory by abort. For participants with capacity larger than the limit, the global arbitration chooses to abort the youngest participant which has the largest participant id. This helps to let the old queries to run to completion. The abort capacity limit is reduced by half if could not find a victim participant until this reaches to zero.

  NOTE: the limit value must be either zero, or a power of 2.

shared-arbitrator.memory-pool-min-reclaim-bytes¶

• Type: string

• Default value: 128MB

  Specifies the minimum bytes to reclaim from a participant at a time. The global arbitration also avoids reclaiming from a participant if its reclaimable used capacity is less than this threshold. This is to prevent inefficient memory reclaim operations on a participant with small reclaimable used capacity, which could cause a large number of small spilled files on disk.

shared-arbitrator.memory-reclaim-threads-hw-multiplier¶

• Type: string

• Default value: 0.5

  Floating point number used in calculating how many threads to use for memory reclaim execution: hw_concurrency x multiplier. 0.5 is default.

shared-arbitrator.global-arbitration-memory-reclaim-pct¶

• Type: string

• Default value: 10

  If not zero, specifies the minimum amount of memory to reclaim by global memory arbitration as percentage of total arbitrator memory capacity.

shared-arbitrator.global-arbitration-abort-time-ratio¶

• Type: string

• Default value: 0.5

  The ratio used with shared-arbitrator.memory-reclaim-max-wait-time, beyond which global arbitration will no longer reclaim memory by spilling, but instead directly abort. It is only in effect when global-arbitration-enabled is true.

shared-arbitrator.global-arbitration-without-spill¶

• Type: string

• Default value: false

  If true, global arbitration does not reclaim memory by spilling, but only by aborting. This flag is only effective if shared-arbitrator.global-arbitration-enabled is true.

Cache Properties¶

The configuration properties of AsyncDataCache and SSD cache are described here.

async-cache-persistence-interval¶

• Type: string

• Default value: 0s

  The interval for persisting in-memory cache to SSD. Set this to a non-zero value to activate periodic cache persistence.

async-data-cache-enabled¶

• Type: boolean

• Default value: true

  In-memory cache.

async-cache-ssd-gb¶

• Type: integer

• Default value: 0

  The size of the SSD. Unit is in GiB (gibibytes).

async-cache-ssd-path¶

• Type: string

• Default value: /mnt/flash/async_cache.

  The path of the directory that is mounted onto the SSD.

async-cache-max-ssd-write-ratio¶

• Type: double

• Default value: 0.7

  The maximum ratio of the number of in-memory cache entries written to the SSD cache over the total number of cache entries. Use this to control SSD cache write rate, once the ratio exceeds this threshold then we stop writing to the SSD cache.

async-cache-ssd-savable-ratio¶

• Type: double

• Default value: 0.125

  The min ratio of SSD savable (in-memory) cache space over the total cache space. Once the ratio exceeds this limit, we start writing SSD savable cache entries into SSD cache.

async-cache-min-ssd-savable-bytes¶

• Type: integer

• Default value: 16777216

  Min SSD savable (in-memory) cache space to start writing SSD savable cache entries into SSD cache.

  The default value 16777216 is 16 MB.

  NOTE: we only write to SSD cache when both async-cache-max-ssd-write-ratio and async-cache-ssd-savable-ratio conditions are satisfied.

async-cache-persistence-interval¶

• Type: string

• Default value: 0s

  The interval for persisting in-memory cache to SSD. Set this configuration to a non-zero value to activate periodic cache persistence.

  The following time units are supported:

  ns, us, ms, s, m, h, d

async-cache-ssd-disable-file-cow¶

• Type: bool

• Default value: false

  In file systems such as btrfs that support cow (copy on write), the SSD cache can use all of the SSD space and stop working. To prevent that, use this option to disable cow for cache files.

ssd-cache-checksum-enabled¶

• Type: bool

• Default value: false

  When enabled, a CRC-based checksum is calculated for each cache entry written to SSD. The checksum is stored in the next checkpoint file.

ssd-cache-read-verification-enabled¶

• Type: bool

• Default value: false

  When enabled, the checksum is recalculated and verified against the stored value when cache data is loaded from the SSD.

cache.velox.ttl-enabled¶

• Type: bool

• Default value: false

  Enable TTL for AsyncDataCache and SSD cache.

cache.velox.ttl-threshold¶

• Type: string

• Default value: 2d

  TTL duration for AsyncDataCache and SSD cache entries.

  The following time units are supported:

  ns, us, ms, s, m, h, d

cache.velox.ttl-check-interval¶

• Type: string

• Default value: 1h

  The periodic duration to apply cache TTL and evict AsyncDataCache and SSD cache entries.

Exchange Properties¶

exchange.http-client.request-data-sizes-max-wait-sec¶

• Type: integer

• Default value: 10

  Maximum wait time for exchange request in seconds.

Memory Checker Properties¶

The LinuxMemoryChecker extends from PeriodicMemoryChecker and is used for Linux systems only. The LinuxMemoryChecker can be enabled by setting the CMake flag PRESTO_MEMORY_CHECKER_TYPE=LINUX_MEMORY_CHECKER. The following properties for PeriodicMemoryChecker are as follows:

system-mem-pushback-enabled¶

• Type: boolean

• Default value: false

If set to true, starts memory limit checker to trigger memory pushback when server is under low memory pressure.

system-mem-limit-gb¶

• Type: integer

• Default value: 60

Specifies the system memory limit that triggers the memory pushback or heap dump if the server memory usage is beyond this limit. A value of zero means no limit is set. This only applies if system-mem-pushback-enabled is true. Set system-mem-limit-gb to be greater than or equal to system-memory-gb but not higher than the available machine memory of the deployment. The default value of 60 gb is calculated based on available machine memory of 64 gb.

system-mem-shrink-gb¶

• Type: integer

• Default value: 8

Specifies the amount of memory to shrink when the memory pushback is triggered. This only applies if system-mem-pushback-enabled is true.

Environment Variables As Values For Worker Properties¶

This section applies to worker configurations in the config.properties file and catalog property files only.

The value in a key-value pair can reference an environment variable by using a leading $ followed by enclosing the environment variable name in brackets ({}).

key=${ENV_VAR_NAME}

The environment variable name must match exactly with the defined variable.

This allows a worker to read sensitive data such as access keys from an environment variable rather than having the actual value hard coded in a configuration file on disk, improving the security of deployments.

For example, consider the hive connector’s hive.s3.aws-access-key property. This is sensitive data and can be stored in an environment variable such as AWS_S3_ACCESS_KEY which is set to the actual access key value.

One mechanism is to create a preload library that is injected at the time presto_server is started that decrypts encrypted secrets and sets environment variables specific to the presto_server process. These can then be referenced in the properties.

Once decrypted the preloaded library sets the AWS_S3_ACCESS_KEY environment variable which then can be accessed by providing it in the catalog properties:

hive.s3.aws-access-key=${AWS_S3_ACCESS_KEY}