Skip to content

Latest commit

 

History

History
63 lines (58 loc) · 13.5 KB

configuration.md

File metadata and controls

63 lines (58 loc) · 13.5 KB

Global Options (System properties / Java -Dxxx)

name description
"CeresDB.available_cpus" Specify the number of CPU cores available to the SDK, and use all the number of CPUs in the current environment by default
"CeresDB.grpc.conn.failures.reset_threshold" Threshold gRPC reset connection, default is 3
"CeresDB.client.read.write.rw_logging" Simple log output switch for read and write, default is true
"CeresDB.client.write.collect_wrote_detail" Whether to keep the written table list after the write is successful, default is false
"CeresDB.client.use_os_signal" Whether to use OS Signal, SIGUSR2 is used by default
"CeresDB.signal.out_dir" The Signal receiver outputs the content to the specified disk directory, which defaults to the process startup directory
"CeresDB.rpc.write.limit_percent" RPC adaptive limit write ratio (default 0.7), the rest is reserved for read
"CeresDB.reporter.period_minutes" Metrics reporter timing output period,default is 30 minutes

CeresDBOptions

name description
clusterAddress The address of ceresDB server, which can be a stand-alone address or a cluster address
routeMode The route mote,support DIRECT and PROXY. DIRECT mode is directly connected to the server, and the routing information is cached on the SDK side, which has better performance; PROXY mode is forwarded through the PROXY module on the server side, which contains performance loss, but it can solve some case that cannot be directly connected, such as network isolation
asyncWritePool A write scheduling thread pool is required in an async client, which can be configured by user,default is SerializingExecutor,If you configure the pool, please focus on the performance metric async_write_pool.time
asyncReadPool Same as asyncWritePool for reading, performance metric is async_read_pool.time
database The database used for client, can be overridden by RequestContext
rpcOptions The configuration options of the rpc, please refer to RpcOptions for details
routerOptions The configuration options of the routing table, please refer to RouterOptions for details
writeOptions The configuration options of the writing, please refer to WriteOptions for details

WriteOptions

name description
maxRetries The max retries for write request, if the server returns an error code, the SDK will determine whether to retry the request; the retry process is transparent to the user, completely asynchronous, and imperceptible to the upper layer
maxWriteSize Ihe maximum data points for each write request, if exceeds, it will be divided into multiple requests, default 512
maxInFlightWritePoints If the maximum number of data points requested in one write request exceeds the current limit, the request will be blocked
limitedPolicy The write limiting policy, provide several implementations is blocking, discard and blocking-timeout,default is abort-blocking-timeout(3s) (Block until timeout 3s and fail with an exception),Users can also extend the policy

QueryOptions

name description
maxRetries Same as WriteOptions.maxRetryies for query
maxInFlightQueryRequests Same as WriteOptions.maxInFlightWriteRows for query
limitedPolicy The query limiting policy, provide implementations smae as WriteOptions.limitedPolicy,but default is abort-blocking-timeout(10s)

RpcOptions

name description
defaultRpcTimeout The rpc call timeout, each request can specify timeout separately, if not specified, default is 10s
rpcThreadPoolSize The maximum number of threads in the thread pool that handles RPC calls
rpcThreadPoolQueueSize The maximum number of threads in pool for processing RPC calls, IO threads will be used to process the response if set to 0, default is 64. The core_size of the thread pool should be smaller than this value, the formula is: Math.min(Math.max(Cpus.cpus() << 1, 16), rpcThreadPoolQueueSize)
maxInboundMessageSize The maximum number of bytes that can be accepted for one Inbound message, default is 64M
flowControlWindow Flow control based on http2.0, default is 64M
idleTimeoutSeconds The maximum idle time of the connection, the connection may fail if the time exceeds, default is 5 minutes
keepAliveTimeSeconds The time interval for sending keep-alive ping on the transport (default is infinite seconds, same as turning off), contain exponential avoidance strategy
keepAliveTimeoutSeconds The timeout for waiting keep-alive ping (default is 3 senconds), if no ack is received within this time, it will close the connection.
keepAliveWithoutCalls If enale this param (defaul is disable), keep-alive pings can be sent even if no requests are made. It has overhead, it is recommended to use idleTimeoutSeconds instead of this option
limitKind Limiting algorithm type, supports Vegas and Gradient, default is Gradient
initialLimit Limiter initial limit
maxLimit Limiter max limit
smoothing Limiter smoothing factor, default 0.2
blockOnLimit Block the Request when limit is exceeded, default is false

RouterOptions

name description
maxCachedSize The maximum number of local cached routing table, default is 10_000, it will be periodically GC if exceeded
gcPeriodSeconds The periodic interval of GC which will clear unused router, default is 60 seconds
refreshPeriodSeconds The periodic interval of refreshing routing table in background, default is 30 seconds.