Why does the WAL directory swell?

Question

PostgreSQL

Configuration.
master - slave
When a rather large amount of data arrives at the master, a bunch of wal files are generated and successfully go to the slave.
at the same time, at some point in time, the slave stops "playing" wal and its directory is full, therefore the slave crashes.
It seems like i \ o are high, but at the same time, the response speed does not increase + the CPU is not in the iowait state. i.e. for the first time

Answer 1

If the replica suspends replication (it can be seen by waiting in the list of processes for the startup process), it means that transactions running on it interfere with it. Check your max_standby_streaming_delay.
In general, why the database can not delete WAL beyond the necessary:
​​- a replication slot is created but it is not read
- archive_command is executed with an error