It's intriguing me that an upgrade from 7.0.5 to 7.0.8 has reduced the usual number of SLOWLOG entries in 2 real world applications we have. For one of our clusters we only get it for a PSYNC, and in another one, it's now very rare, not daily like before.
I wonder if there was an optimization in this upgrade, if it's a bugfix or an actual bug.
I see things like https://github.com/redis/redis/pull/11541, but 5% improv shouldn't be responsible for wiping our SLOWLOG like this.
Our usual top commands are evalsha, hget, sunion, smembers, hgetall and sort_ro Ideas?
Thank you
Comment From: filipecosta90
Our usual top commands are evalsha, hget, sunion, smembers, hgetall and sort_ro
@eduardobr can you provide us with the info commandstats and info latencystats of 7.0.5 and 7.0.8 for your usage pattern?
Furthermore, if possible, we would love to have more detail about the usage patterns of those commands ( specifically evalsha ) -- can you describe them ( assuming you can :) )?
Comment From: eduardobr
@filipecosta90, unfortunately I can't get data for 7.0.5 anymore. About evalsha, it's used for reading data: input 2 small strings, get a bool response.