The discover performance measure problems and analyze


To analyze latency problems re DIS, can the built-in latency tool by REDIS-cli use, with the order 100 Requests per second on the re DIS be pinged and it can parse the response time.

Consoles BSP:

In the interval i = 10 seconds on localhost tested

redis-cli --latency-history -i 10 -h 127.0.0.1 -p 6379

Output:

min: 0, max: 10, avg: 0.89 (906 samples) -- 10.01 seconds range
min: 0, max: 32, avg: 1.44 (862 samples) -- 10.00 seconds range
min: 0, max: 6, avg: 0.50 (939 samples) -- 10.00 seconds range
min: 0, max: 7, avg: 0.47 (941 samples) -- 10.01 seconds range
min: 0, max: 12, avg: 0.47 (942 samples) -- 10.00 seconds range
min: 0, max: 137, avg: 2.70 (779 samples) -- 10.01 seconds range
min: 0, max: 119, avg: 5.92 (626 samples) -- 10.06 seconds range
min: 0, max: 2389, avg: 7.31 (572 samples) -- 10.00 seconds range
min: 0, max: 14, avg: 0.64 (926 samples) -- 10.00 seconds range
min: 0, max: 9, avg: 0.50 (938 samples) -- 10.01 seconds range
min: 0, max: 16, avg: 0.47 (941 samples) -- 10.01 seconds range
min: 0, max: 21, avg: 0.56 (933 samples) -- 10.01 seconds range
min: 0, max: 23, avg: 0.56 (933 samples) -- 10.01 seconds range
min: 0, max: 150, avg: 1.95 (827 samples) -- 10.01 seconds range
min: 0, max: 131, avg: 4.22 (696 samples) -- 10.01 seconds range
min: 0, max: 131, avg: 7.06 (581 samples) -- 10.01 seconds range
min: 0, max: 4136, avg: 10.33 (488 samples) -- 10.01 seconds range
min: 0, max: 17, avg: 1.15 (886 samples) -- 10.02 seconds range
min: 0, max: 18, avg: 0.88 (906 samples) -- 10.01 seconds range
min: 0, max: 11, avg: 0.61 (928 samples) -- 10.00 seconds range
min: 0, max: 12, avg: 0.55 (936 samples) -- 10.01 seconds range
min: 0, max: 9, avg: 0.51 (937 samples) -- 10.01 seconds range

Here you can see well, that the problems at a distance of 60 Occur seconds. This was because in the case of, that re DIS has persisted is every 60 s on the hard disk. The corresponding option in the redis.conf was:

save 60 1000

and with the option

save ""

may that persist on the hard disk are turned off.

 

But also the option

appendonly yes

had on “no” be set, because there was this on too many writes.