Slave in replication does not keep up with the master if it is far behind

Question

MySQL

MySQL 5.6.13
~200 requests per second
ROW-Based

Replication when the difference in data is small, it quickly catches up with the master and then everything is fine.
If the slave is stopped for a couple of days, and then started again, then it starts to catch up for about two minutes very cheerfully (high processor load, I / O, everything is as it should be). And then it practically falls asleep - the processor load is 1%, there is little I / O, Seconds_Behind_Master decreases very slowly, stands still or even increases.

In terms of performance, the hardware of the slave is higher, the software versions are identical. There are no errors in the logs. Between gigabit servers.

Where to dig?

Answer 1

seems to have figured out
the problem in bcache - merged the database onto a clean SSD and everything flew off.
very strange
, %iowait remained approximately the same - 5.4
bcache may have problems with multi-threaded access, because everything in synthetics is chocolate.
in any case, the question is closed, MySQL is not to blame :)

Answer 2

1. Check if replication is working correctly. That is, did some error SHOW SLAVE STATUS\G appear in the Last_Error variable
2. Check the age of the binary logs on the master. They might be out in 2 days.