During indexer restart/ indexer cluster rolling restart TcpInputProcessor fails to drain queue.

[hrawat]

During graceful indexer/HF restart/stop (basically where ever splunktcp is configured) if you see last entries in metrics.log before splunk finally stops. 
Where splunktcpin queue (name=splunktcpin) shows current_size, largest_size, smallest_size has same value( but parsingqueue to indexqueue none blocked), TcpInputProcessor fails to drain splunktcpin queue despite parsingqueue or indexqueue are empty. 

02-18-2024 00:54:28.370 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1507, largest_size=1507, smallest_size=1507
02-18-2024 00:54:28.370 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=indexqueue, max_size_kb=10240, current_size_kb=7, current_size=40, largest_size=40, smallest_size=0
02-18-2024 00:54:28.368 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1148, largest_size=1148, smallest_size=1148
02-18-2024 00:54:28.368 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=indexqueue, max_size_kb=10240, current_size_kb=7, current_size=40, largest_size=40, smallest_size=0
02-18-2024 00:53:57.364 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1507, largest_size=1507, smallest_size=1507
02-18-2024 00:53:57.364 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:53:57.362 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1148, largest_size=1148, smallest_size=1148
02-18-2024 00:53:57.362 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:53:26.372 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1507, largest_size=1507, smallest_size=1507
02-18-2024 00:53:26.372 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:53:26.370 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1148, largest_size=1148, smallest_size=1148
02-18-2024 00:53:26.370 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:52:55.371 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1507, largest_size=1507, smallest_size=0
02-18-2024 00:52:55.371 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:52:55.369 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=splunktcpin, blocked=true, max_size_kb=500, current_size_kb=499, current_size=1148, largest_size=1148, smallest_size=0
02-18-2024 00:52:55.369 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:52:24.397 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=splunktcpin, max_size_kb=500, current_size_kb=0, current_size=0, largest_size=30, smallest_size=0
02-18-2024 00:52:24.396 +0000 INFO  Metrics - group=queue, ingest_pipe=1, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0
02-18-2024 00:52:24.380 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=splunktcpin, max_size_kb=500, current_size_kb=0, current_size=0, largest_size=16, smallest_size=0
02-18-2024 00:52:24.380 +0000 INFO  Metrics - group=queue, ingest_pipe=0, name=indexqueue, max_size_kb=10240, current_size_kb=0, current_size=0, largest_size=1, smallest_size=0

During graceful shutdown pipeline processors are expected to drain the queue.
This issue is fixed in 9.2.1 and 9.1.4. 

[gjanders]

This looks very useful, is there a recommended way to set the maxSendQSize ?

Do I need to vary it depending on the thruput of the HF per pipeline?

I'm assuming the maxSendQSize would be in-memory buffer/queue per-pipeline in addition to the overall maxQueueSize?

Finally I'm assuming this would be useful when there is no load balancer in front of the indexers?

[tej57]

Thank you for the insights @hrawat .

I believe this should be part of Monitoring Console as well to identify the queue behavior.

Thanks,
Tejas.