Closed vponomaryov closed 2 years ago
Latest state of this PR was tested here: https://jenkins.scylladb.com/view/staging/job/scylla-staging/job/valerii/job/longevity-large-partition-asymmetric-cluster-3h/40/consoleFull It works. Example of output of the failed loader:
1h53m16.9s 5589 55890 0 24ms 24ms 23ms 22ms 22ms 21ms 18ms
1h53m17.9s 5597 55970 0 25ms 24ms 23ms 22ms 22ms 21ms 18ms
1h53m18.9s 5591 55910 0 30ms 27ms 23ms 23ms 22ms 21ms 18ms
2022/11/04 18:24:43 EOF
1h53m19.6s 5568 55680 0 30ms 29ms 24ms 22ms 22ms 21ms 18ms
Results
Time (avg): 1h53m19.580764851s
Total ops: 33454696
Total rows: 334535935
Total errors: 1
Operations/s: +Inf
Rows/s: +Inf
raw latency :
max: 50.593791ms
99.9th: 31.064063ms
99th: 23.789567ms
95th: 22.544383ms
90th: 22.151167ms
median: 20.971519ms
mean: 20.320071ms
c-o fixed latency :
max: 50.593791ms
99.9th: 31.064063ms
99th: 23.789567ms
95th: 22.544383ms
90th: 22.151167ms
median: 20.971519ms
mean: 20.323227ms
Following critical errors were caught during the run:
Error limit (maxErrors) of 1 errors is reached
Error limit (maxErrorsAtRow) of 1 errors is reached
Now it is possible to specify number of errors after which a stress runner will stop the execution. To do it just add following cmd option to the s-b command:
-error-limit n
If it is not set it gets default value as0
which meansno limit
(status quo).