I see that druid ingestion task is stuck (99.9…% finished loading. It is written in the output log again and again).
The process consumes more and more memory but nothing changes. In the end it crashes because of memory issues.
Then there is no way to recover. Every ingestion task fails. I have to remove the whole var directory and start from scratch.
Is there any better solution to that situation?
What type of ingestion is this? If this is during ingestion phase, try to reconfigure tunig config for the ingestion to persist often or reduce the number of rows per segment
If it is during publishing phase then review middlemanager config.
Thanks & Rgds
well I found out it happens when there is no place in cache. The cache should be bigger or some segments should be mark unused.