Created 12-14-2021 08:07 AM
Hello Team,
We are facing Slow ReadProcessor warnings while pulling data from kafka with spark applications. After few slow ReadProcesser warnings, the applications fail. A partial log is attached. Please let us know if you need further information.
Please find below warning message,i am frequently i am seeing this logs and also my application taking too long to complete.
2021-12-13 03:25:00 WARN DFSClient:854 - Slow ReadProcessor read fields took 117390ms (threshold=30000ms); ack: seqno: 353 reply: SUCCESS reply: SUCCESS reply: SUCCESS downstreamAckTimeNanos: 778712 flag: 0 flag: 0 flag: 0, targets: [DatanodeInfoWithStorage[10.108.0.18:1019,DS-ec5cff3e-e958-416e-9ad8-de319cfbc28a,DISK], DatanodeInfoWithStorage[10.108.0.106:1019,DS-61163e3d-59ef-4dfc-b194-7385cff86a7c,DISK], DatanodeInfoWithStorage[10.108.0.96:1019,DS-af490217-ef46-4d92-bd6e-78bda82c82dc,DISK]]
Thanks & Regards
Kallem
Created 12-14-2021 04:08 PM
Couple of possibilities for this WARN messages are:
1) If there is any GC issue on the datanode, this type of WARN messages is seen.
2) If there is any disk issue
3) the last possibility is network latency/slowness between the application, Kafka node, and datanode.
Created 12-15-2021 04:59 PM
I agree @Nandinin's suggestion. Adding some thoughts on hdfs side for your reference:
1. Now you know which 3 DNs maybe slow in the pipeline and the timestamp. So you can go to each datanode log, to see if there are "JvmPauseMonitor" ? or "Lock held"? or other WARN / ERROR ?
2. Refer to this KB https://my.cloudera.com/knowledge/Diagnosing-Errors-Error-Slow-ReadProcessor-Error-Slow?id=73443, check the Slow message from DN logs around the above timestamp to determine what is the main cause.
Regards,
Will
Created 12-20-2021 10:45 AM
@Kallem Has the reply helped resolve your issue? If so, please mark the appropriate reply as the solution, as it will make it easier for others to find the answer in the future. Thanks!
Regards,
Diana Torres,