AOF fsync policies, possible to improve latency/stall times #700

Digitalroute opened this Issue Oct 8, 2012 · 0 comments

1 participant


When doing a rewrite of large AOF files it should be possible to do extra fsync during the write, this since doing on big fsync in the end can block the main process for a long time and lead to high stall times/ latency.

We have a patch for this porblem:

What this patch includes is a new property in the redis.conf that can be set to do extra fsyncs based on how many objects have been handled from current database.

Seems to be similar to this problem concerning the .rdb files:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment