Redis Sentinel: When it rewrites sentinel.conf it permanetely deletes parallel-sync, down-after-milliseconds and failover-timeout

Environment:

Debian Wheezy Redis 2.8.6 (official backport version)

Confirmed by:

@badboy (badboy_ on IRC) said it also affects the unstable branch.

Setup:

• A sentinel instance
• A master server instance
• A client server instance that happens to be on the sentinel instance
• Probably unrelated but maybe worth mentioning

When I redis-cli into the sentinel and issue a failover command, it reports back OK as expected.

When I open sentinel.conf on the sentinel instance the IPs are correctly adjusted:

It also adds these commands: - epoch - known-slave

It retains both script commands.

The issue:

It completely removes/strips: - parallel-sync - down-after-milliseconds - failover-timeout

Comment From: mattsta

Do you have an example of your original config vs. the rewritten config? Some options are not written if their values match the default value for the option.

Current defaults for your missing lines:

#define SENTINEL_DEFAULT_PARALLEL_SYNCS 1  
#define SENTINEL_DEFAULT_FAILOVER_TIMEOUT (60*3*1000)
#define SENTINEL_DEFAULT_DOWN_AFTER 30000      

Comment From: AntelopeSalad

Yes, those were the values. I had them set as the defaults explicitly in my config.

Comment From: badboy

Ah, that actually makes sense. I had no time to check it correctly but probably my test used default values too

Edit: Had a look at my testing setup. Indeed I used the defaults for parallel syncs and failover timeout. Issue can be closed then.