Shulou(Shulou.com)06/03 Report--

Yarn HA

RM:

a. When starting, it will write a lock file to the / hadoop-ha directory of ZK. If it is successful, it will be active, otherwise standyby.

The RM of standy will always monitor whether the lock file exists. If it does not exist, it will try to create it for active RM.

b. Will receive task requests from the client, receive and monitor nm resource reports, be responsible for resource allocation and scheduling, start and monitor our ApplicationMaster

NM:

Resource management on the node, start the container container, run task calculation, report resources, container status to RM and task processing report to job ApplicationMaster

ZKFC is embedded directly into RM.

RMStateStore is stored directly in ZK

ApplicationMaster (AM):

The management and scheduling of the task of a single application (job), and requests for resources to the rm, sends launch conta instructions to the nm, and receives the processing status information of the task of the NM.

Interview question: where does ApplicationMaster run?

ApplicationMaster container that runs on a nm machine. AM is also a task, just doing its own job.

RMstatestore:

A.RM 's job information is stored under / rmstore of ZK, to which active RM writes job information (that is, Application).

B when activerm dies and another standby rm is successfully converted to activerm, the corresponding job information is read from the / rmstore directory and the memory information of the job is rebuilt.

Then start the internal service, start to accept the heartbeat of NM, build information about cluster resources, and receive requests from clients to submit jobs, and so on.

ZKFC:

Automatic failover starts only as a thread in RM, not as a separate daemon.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.