We have a cluster with many replica. We found that loading image is fast but save image is slow.
We add some logs like follows:
2021-05-11 20:53:56,163 WARN (leaderCheckpointer|93) [Catalog.saveImage():2000] Backends done 2780888362
2021-05-11 21:13:23,941 WARN (leaderCheckpointer|93) [Catalog.saveImage():2002] DB done 2255394013
Profile shows FileOutputStream.writeBytes cost much time
Suggest: Use buffer out put stream to reduce the number of system calls.
We have a cluster with many replica. We found that loading image is fast but save image is slow.
We add some logs like follows:
Profile shows
FileOutputStream.writeBytescost much timeSuggest: Use buffer out put stream to reduce the number of system calls.