Typical Flink Parameters

**Table 1** Typical Flink parameters
Parameter	Description	Recommended Value	Description
-c	Main class name	Set this parameter as needed.	Yes
-yjm	Memory of the JobManager process. The default value is 2 GB.	Set this parameter as needed.	No
-ytm	Memory of the TaskManager process. The default value is 4 GB.	Set this parameter as needed.	No
-ynm	Name of the Flink YARN job.	Set this parameter as needed.	Yes
-ys	Number of slots in the TaskManager.	2	No
execution.checkpointing.interval	Interval for triggering a checkpoint, which can be added using -yD. The unit is ms.	60000	Yes
execution.checkpointing.timeout	Checkpoint timeout interval. You can run the -yD command to add a checkpoint timeout interval. The default value is 30 minutes.	30min	Yes
execution.checkpointing.tolerable-failed-checkpoints	Total number of tolerable checkpoint failures, which can be added using -yD.	1000	No
state.checkpoints.num-retained	Number of reserved checkpoints, which can be added using -yD.	5	No
state.backend	RocksDB used as the state backend, which can be added using -yD.	rocksdb	Enabled
state.backend.incremental	Incremental RocksDB state backends, which can be added using -yD.	TRUE	Yes
state.backend.rocksdb.block.blocksize	Size of data blocks on the write state backend. The value can be added using -yD.	512KB	Yes
state.backend.rocksdb.block.cache-size	Size of the block cache on the state backend, which can be added using -yD.	1024MB	Yes
taskmanager.memory.jvm-overhead.max	Maximum size of the local memory used for other JVM overheads, such as stack space and garbage collection space. The value can be added using -yD.	10g	No
taskmanager.memory.jvm-overhead.fraction	Ratio of the local memory used for other JVM overheads to TM memory, such as stack space and garbage collection space. The value can be added using -yD.	0.2	No
parallelism.default	Job parallelism. For example, to add the job parallelism for the join operator, use -yD. The default value is 1.	Set this parameter as needed.	No
table.exec.state.ttl	TTL (join ttl) of Flink state, which can be added using -yD. The default value is 0.	Set this parameter as needed.	Yes
heartbeat.timeout	Heartbeat timeout interval between the JM and TM, which can be added using -yD.	1800000	Yes
akka.ask.timeout	Akka communication timeout, which can be added using -yD.	240s	Yes
taskmanager.memory.segment-size	Size (in bytes) of the memory buffer block used by the memory management and network stack. The default value is 32768 (32 KB). You can add the value using -yD.	64kb	No
taskmanager.network.memory.max-buffers-per-channel	Maximum number of buffers that each channel can hold. If there are many idle segments, you can set a larger value. Otherwise, the channel will be blocked because it cannot obtain segments. You can run -yD to set this parameter.	100	No
taskmanager.network.memory.buffers-per-channel	Number of buffers exclusively used by each channel, which can be added by using -yD.	10	No
taskmanager.network.memory.floating-buffers-per-gate	Number of floating buffers used by each channel, which can be added by using -yD.	2000	No
taskmanager.network.netty.server.numThreads	Number of Netty server threads in each TaskManager, which can be added using -yD.	20	No
taskmanager.network.netty.client.numThreads	Number of Netty client threads in each TaskManager, which can be added using -yD.	20	No
state.backend.rocksdb.files.open	Maximum number of files that can be opened. The value -1 indicates that there is no limit. You can run -yD to set this parameter.	-1	No
state.backend.rocksdb.compaction.level.use-dynamic-size	Dynamic data volume threshold of data storage at each layer. You can use -yD to set this parameter.	TRUE	No
state.backend.rocksdb.levels.num	Number of compaction layers allowed by RocksDB, which can be added using -yD.	10	No
state.backend.rocksdb.compaction.style	Compaction algorithm, which can be added using -yD.	FIFO	No
state.backend.rocksdb.verify.checksum	Whether to disable data check during read. You can run -yD to set this parameter.	FALSE	No
state.backend.rocksdb.thread.num	Maximum number of concurrent threads responsible for flushing and compaction in the background. You can run -yD to set this parameter.	4	No
state.backend.rocksdb.writebuffer.count	Maximum number of memtables, which can be added using -yD.	5	No
state.backend.rocksdb.writebuffer.number-to-merge	Minimum number of memtables that are merged before flush. The value can be added using -yD.	3	No
state.backend.rocksdb.background.compaction.max	Maximum number of compaction threads, which can be added using -yD.	10	No
state.backend.rocksdb.flush.max	Number of RocksDB flush threads, which can be added using -yD.	1	No