AOF file corruption in Redis 4.0.9

[moweintraub]

We're running into a situation where we are seeing the AOF file get corrupted when running cluster in redis 4.0.9.

Here is a snippet of the output we're seeing on startup:

5880:C 18 May 20:43:11.748 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
5880:C 18 May 20:43:11.748 # Redis version=4.0.9, bits=64, commit=00000000, modified=0, pid=5880, just started
5880:C 18 May 20:43:11.748 # Configuration loaded
5880:M 18 May 20:43:11.750 * Node configuration loaded, I'm cf83a5e5ef506d903c6ce0c6a0825e7a150cc978
5880:M 18 May 20:43:11.751 * Running mode=cluster, port=6700.
5880:M 18 May 20:43:11.751 # WARNING: The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128.
5880:M 18 May 20:43:11.751 # Server initialized
5880:M 18 May 20:43:12.364 # Bad file format reading the append only file: make a backup of your AOF file, then use ./redis-check-aof --fix <filename>

I've tried looking back in the output to see what was logged before this started happening, and will add it to the issue once I'm able to get it.

Thanks,
Mo

[soloestoy]

Do you startup redis with config option rdbchecksum no?

A similar issue had been submitted #4902, and if you meet the same situation, you can set rdbchecksum yes, or try to apply my patch in PR #4908.

[moweintraub]

We have rdbchecksum set to yes. We're also looking into getting more information to add to the ticket.