I monitor my ES cluster merge process through http api /_nodes/stats/indices every 1 min and record the diff between.
Here is my sampled data:
Given 2 data collected
18:00 --> total_size_in_bytes: 145600300; total_time_in_millis: 23000
18:01 --> total_size_in_bytes: 195000000; total_time_in_millis: 25000
- Can I say about 50 MB merged and cost 2 sec
If the answer is yes, 50MB is estimated "write 50MB" to disk as final merged segment file or read 50MB from other segment files those are being merged? If this is "disk write", the performance is 25MB/per sec? Can I throttle this behavior?
Your help is highly appreciated. Thank.