[CASSANDRA-5896] Stress reports invalid latencies

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 2.1 beta1
• Component/s: Legacy/Tools
• Severity: Normal
• Since Version: 2.1 rc3
• Assignee: David Brosius
• Reporter: Ryan McGuire

Authors

David Brosius

Reviewers

Jonathan Ellis

• Created: 15/Aug/13 17:44
• Updated: 16/Apr/19 09:32
• Resolved: 16/Aug/13 12:40

Description

I get this output from trunk:

$ ccm node1 stress
total,interval_op_rate,interval_key_rate,latency,95th,99th,elapsed_time
176678,17667,17667,153510.0,1322730.2,181065524.2,10
335566,15888,15888,145371.0,1220290.5,181065524.2,20
498488,16292,16292,145535.0,1227675.0,46032966.9,30
642012,14352,14352,143999.0,1130215.4,46004055.9,40
776083,13407,13407,145573.5,1737871.2,153211818.5,50
1000000,22391,22391,145128.0,1336373.6,148773096.0,60


Averages from the middle 80% of values:
interval_op_rate          : 15521
interval_key_rate         : 15521
latency median            : 146797.7
latency 95th percentile   : 1327756.5
latency 99.9th percentile : 121475978.0
Total operation time      : 00:01:00
END

Notice the wild latency values.

Whereas this is the output that I've come to expect (from cassandra-2.0):

$ ccm node1 stress
total,interval_op_rate,interval_key_rate,latency,95th,99th,elapsed_time
157972,15797,15797,0.1,1.2,119.1,10
346627,18865,18865,0.1,1.2,118.8,20
493937,14731,14731,0.1,1.2,119.2,30
663086,16914,16914,0.2,1.3,119.8,40
893083,22999,22999,0.1,1.1,65.6,50
1000000,10691,10691,0.1,1.1,65.6,55


Averages from the middle 80% of values:
interval_op_rate          : 17861
interval_key_rate         : 17861
latency median            : 0.1
latency 95th percentile   : 1.2
latency 99.9th percentile : 108.5
Total operation time      : 00:00:55
END