You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, flink-sql-gateway can only submit job with a specified executor which defined in flink-conf.yaml。But if someone wants to submit a job to yarn-session, and others want to do it with standalone cluster, then flink-sql-gateway can't satisfy this right now. So could flink-sql-gateway support this feature future。
And I also has a roughly idea that how flink-sql-gateway supports this feature. we can introduce a new component which I called ExecutionManager .ExecutionManager is like SessionManager, but it's main function is to creating a new Execution, each Execution represents an deploy mode(yarn-seesion, yarn-per-job, standalone, local, etc). The Execution will create a new SessionManager and will use user specified execution.target to instance a DefaultContext. By this way, every new session created by SessionManager can submit job to different cluster.
Furthermore, flink-sql-gateway do not need to create SessionManager at start time but instead of ExecutionManager, and the SessionManager will be instance by ExecutionManager.
And, we'll must implement a new netty handler that used to create/delete/get Execution.
The text was updated successfully, but these errors were encountered:
Currently, flink-sql-gateway can only submit job with a specified executor which defined in flink-conf.yaml。But if someone wants to submit a job to yarn-session, and others want to do it with standalone cluster, then flink-sql-gateway can't satisfy this right now. So could flink-sql-gateway support this feature future。
And I also has a roughly idea that how flink-sql-gateway supports this feature. we can introduce a new component which I called
ExecutionManager
.ExecutionManager
is likeSessionManager
, but it's main function is to creating a newExecution
, eachExecution
represents an deploy mode(yarn-seesion, yarn-per-job, standalone, local, etc). TheExecution
will create a newSessionManager
and will use user specifiedexecution.target
to instance aDefaultContext
. By this way, every newsession
created bySessionManager
can submit job to different cluster.Furthermore, flink-sql-gateway do not need to create
SessionManager
at start time but instead ofExecutionManager
, and theSessionManager
will be instance byExecutionManager
.And, we'll must implement a new netty handler that used to create/delete/get
Execution
.The text was updated successfully, but these errors were encountered: