Determining why Redis is receiving a SIGTERM every minute or two

Question

I'm running redis-server on Ubuntu 16.04. After a recent upgrade to redis 3.2.7 I now see lines like

signal-handler (1486427812) Received SIGTERM scheduling shutdown... every minute or two in the redis log and then redis restarts.

The only other suspicious line in the logs is:

# WARNING: The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128.

Any advice on how to diagnose why this is happening would be greatly appreciated. Thanks!

Answer 1

For anyone else still looking for a solution, Type=forking (which is default for redis installed with apt) is incompatible with supervised=systemd in redis.conf.

It has to be updated to Type=simple in /etc/systemd/system/redis.service.

Forking is only required in daemonized mode, but systemd managed redis overrules the daemonized setting.

Change from Restart=always to Restart=on-failure

For reference- redis.conf, https://www.freedesktop.org/software/systemd/man/systemd.service.html

Answer 2

Just for the record: As in J G Miller's comment, you have to watch that "Type=simple" and "daemonize yes" settings are incompatible. I struggeld a few days while I found this comment. My redis just simply gave the start-limit-hit error because when the service started it immediately got the SIGTERM signal and stopped.

So either change in the service file the "Type=forking" or the "daemonize no" lines to solve this, depending on your needs.