redis-benchmark tool (version 7.0.2) gives the following error.
# redis-benchmark -t set -n 10000 -q -h <IP_address> -p <Port> --csv -r 1000 --cluster -c 10 -d 3000 -l
Cluster has 5 master nodes:
.
.
"test","rps","avg_latency_ms","min_latency_ms","p50_latency_ms","p95_latency_ms","p99_latency_ms","max_latency_ms"
Error: Protocol error, got "\r" as reply type byte
It appears to be a bug. Let us know the fix or work around.
Comment From: madolson
@mukundv18 Can you provide more information about your setup, I wasn't able to trivially reproduce it on Redis 7.0.2. Please include the configs if possible.
The error looks like it's trying to read the type, but failing and just throwing an error.
Comment From: mukundv18
@madolson Thanks! Found the issue. The AOF file had become too big that there was no space left. We are collecting both RDB and AOF. Can you suggest a config setting that ensures that the AOF file collects data only from the time of the last Snapshot? If configured that way, are there any use cases which we need to be aware of? Thanks!
Comment From: madolson
@mukundv18 The way you mentioned is how the AOF works, it accumulates changes up until a certain threshold is met and then triggers a rewrite. See auto-aof-rewrite-percentage and auto-aof-rewrite-min-size for configuring that threshold.
Also note that we try to keep the issues tab of the project for bugs in the project, if you want help with running Redis, I would suggest checking out the links here: https://redis.io/community/.
Comment From: oranagra
i fail to see the connection between the protocol error returned to redis-benchmark and the disk space issue for AOF. maybe it could have been this: #10786 which caused protocol error when redis attempts to respond with an error after being unable to persist, but that bug was fixed in 7.0.1.
if there's another bug causing protocol error, let's re-open this and figure out how to reproduce it.