I know what PostgreSQL checkpoint is and when it is happening.
I need some additional information about the logs produced by the log_checkpoints = on
parameter, so please explain some points of it to me:
2017-09-09 16:31:37 EEST [6428-6524] LOG: checkpoint complete: wrote 30057 buffers (22.9%); 0 transaction log file(s) added, 0 removed, 47 recycled; write=148.465 s, sync=34.339 s, total=182.814 s; sync files=159, longest=16.143 s, average=0.215 s
- I know that 22.9% of shared buffers are written (I have 1024 MB
shared_buffers
so that means 234 MB are written out). - I know that 47 WAL files are recycled, i.e., they are not needed anymore for crash recovery, because the real data from them is already on disk.
Question A. But what about write=148.465 s
and sync=34.339
? What is the difference? What is write
and why its time is far more than the fsync()
operation?
Question B. What are sync files
? Which files: WAL files? Why sync files
are 159, but there are only 47 recycled files? What is the relation between these?
Thank you!