JanusGraph full GC doesn't recycle heap memory
Raul Guo <zaixiagu...@...>
I am trying to save data to my JanusGraph(Hbase backend, with no mixed index backend) concurrently(with 100 thread bounded transactions to do with requests), I found a very strange issue, please give me a hand.
When there are many requests at the beginning, the JVM heap is soon exausted, (I have disabled the database level cache, and set
GC with little effect
I tried to find out what is taking up the memories and refused to be collected, so I used the JMap to print the memory object allocation, the top result is listed here:
It looks very like the cache data, but I have disabled the database level cache, and set the vertex cache to very low value, so where is those objects listed above from? And how should I do with it? If the used memory cannot be recycled, it will be used up sooner or later. Any advice is grateful, thanks.