[HDFS-14186] blockreport storm slow down namenode restart seriously in large cluster

• Type: Improvement
• Status: Reopened
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.7.1
• Component/s: namenode
• Assignee: Xiaoqiao He
• Reporter: Xiaoqiao He
• Created: 05/Jan/19 12:30
• Updated: 18/Apr/20 09:26

Description

In the current implementation, the datanode sends blockreport immediately after register to namenode successfully when restart, and the blockreport storm will make namenode high load to process them. One result is some received RPC have to skip because queue time is timeout. If some datanodes' heartbeat RPC are continually skipped for long times (default is heartbeatExpireInterval=630s) it will be set DEAD, then datanode has to re-register and send blockreport again, aggravate blockreport storm and trap in a vicious circle, and slow down (more than one hour and even more) namenode startup seriously in a large (several thousands of datanodes) and busy cluster especially. Although there are many work to optimize namenode startup, the issue still exists.
I propose to postpone dead datanode check when namenode have finished startup.
Any comments and suggestions are welcome.

Issue Links

• is related to: HADOOP-12173 NetworkTopology#add calls NetworkTopology#toString always
• is related to: HDFS-9198 Coalesce IBR processing in the NN

Sub-Tasks

• Make queue size of BlockReportProcessingThread configurable: Resolved
• Avoid to process duplicate blockreport of datanode when namenode in startup safemode: Resolved
• Avoid block report retry and slow down namenode startup: Open
• Optimizing safemode leave mechanism: Open