If the master redis instance that a redis database is slaved to is down or inaccessible the slave writes:
I/O error writing to MASTER: Connection refused
to the system log every second. This quickly results in a lot of log lines and can mask other data going to the system log.
After writing it once it should not repeat it while the problem continues (except perhaps for a reminder every few minutes).
You might also want to consider if it should also reduce the frequency that it retries the connexion once it has failed for a time.