Summary: | Summariser : Active thread count in is not correct during high load distributing test when using statistical mode | ||
---|---|---|---|
Product: | JMeter - Now in Github | Reporter: | Liu XP <liu_xp2003> |
Component: | Main | Assignee: | JMeter issues mailing list <issues> |
Status: | NEW --- | ||
Severity: | normal | CC: | p.mouawad |
Priority: | P2 | ||
Version: | 3.1 | ||
Target Milestone: | --- | ||
Hardware: | PC | ||
OS: | All |
Description
Liu XP
2017-07-06 01:51:30 UTC
Hello, Thanks for report. Can you test with default mode (MODE_STRIPPED_BATCH) StrippedAsynch. Thank you When used default mode (MODE_STRIPPED_BATCH) StrippedAsynch, it will used up network bandwith on JMeter client side and the count of active thread number also not correct. The below is part of logs: 2017/07/21 15:03:04 INFO - jmeter.samplers.BatchSampleSender: Using batching (client settings) for this run. Thresholds: num=100, time=5000 2017/07/21 15:03:04 INFO - jmeter.samplers.DataStrippingSampleSender: Using DataStrippingSampleSender for this run 2017/07/21 15:03:04 INFO - jmeter.JMeter: Started remote host: 172.18.50.184 (1500620584062) 2017/07/21 15:03:04 INFO - jmeter.engine.ClientJMeterEngine: sent test to 172.18.51.15 basedir='.' 2017/07/21 15:03:04 INFO - jmeter.engine.ClientJMeterEngine: Sending properties {sServerIP=172.18.2.11, nSize=1, nThreadNum=5000, nRampUp=5, nSleepTimes=0.25, nDuration=180} 2017/07/21 15:03:04 INFO - jmeter.engine.ClientJMeterEngine: sent run command to 172.18.51.15 2017/07/21 15:03:04 INFO - jmeter.engine.DistributedRunner: Remote engines have been started 2017/07/21 15:03:04 INFO - jmeter.JMeter: Started remote host: 172.18.51.15 (1500620584200) 2017/07/21 15:03:10 INFO - jmeter.reporters.Summariser: summary + 13592 in 00:00:06 = 2229.3/s Avg: 261 Min: 249 Max: 612 Err: 0 (0.0000%) Active: 29318 Started: 26887 Finished: 0 2017/07/21 15:03:10 INFO - jmeter.reporters.Summariser: summary = 13593 in 00:00:07 = 2064.2/s Avg: 261 Min: 249 Max: 612 Err: 0 (0.0000%) 2017/07/21 15:03:21 INFO - jmeter.reporters.Summariser: summary + 22202 in 00:00:12 = 1874.2/s Avg: 295 Min: 249 Max: 962 Err: 0 (0.0000%) Active: 33402 Started: 30971 Finished: 0 2017/07/21 15:03:21 INFO - jmeter.reporters.Summariser: summary = 35795 in 00:00:18 = 1942.1/s Avg: 282 Min: 249 Max: 962 Err: 0 (0.0000%) 2017/07/21 15:03:31 INFO - jmeter.reporters.Summariser: summary + 7254 in 00:00:10 = 730.0/s Avg: 309 Min: 249 Max: 950 Err: 0 (0.0000%) Active: 33402 Started: 30971 Finished: 0 2017/07/21 15:03:31 INFO - jmeter.reporters.Summariser: summary = 43049 in 00:00:28 = 1517.4/s Avg: 287 Min: 249 Max: 962 Err: 0 (0.0000%) 2017/07/21 15:03:43 INFO - jmeter.reporters.Summariser: summary + 2184 in 00:00:11 = 191.2/s Avg: 291 Min: 250 Max: 712 Err: 0 (0.0000%) Active: 33402 Started: 30971 Finished: 0 2017/07/21 15:03:43 INFO - jmeter.reporters.Summariser: summary = 45233 in 00:00:40 = 1136.7/s Avg: 287 Min: 249 Max: 962 Err: 0 (0.0000%) 2017/07/21 15:04:00 INFO - jmeter.reporters.Summariser: summary + 2233 in 00:00:17 = 130.3/s Avg: 295 Min: 250 Max: 633 Err: 0 (0.0000%) Active: 33402 Started: 30971 Finished: 0 2017/07/21 15:04:00 INFO - jmeter.reporters.Summariser: summary = 47466 in 00:00:57 = 833.6/s Avg: 287 Min: 249 Max: 962 Err: 0 (0.0000%) ... Hello, So for your last test, your issue looks to me like a consequence of a saturated network right ? For initial test, did you check that network between jmeter servers and client is not saturated ? Did you add a unique ID to Thread Group names of each thread ? So for your last test, your issue looks to me like a consequence of a saturated network right ? Yes. For initial test, did you check that network between jmeter servers and client is not saturated ? I had checked network bandwith in initial test. The network between jmeter servers and client is not saturated because I used Statistical mode and set num_sample_threshold=1000 to reduce network bandwidth. Did you add a unique ID to Thread Group names of each thread ? I hadn't added unique ID to Thread Group name because I don't know where to trace them with Statistical mode. This issue has been migrated to GitHub: https://github.com/apache/jmeter/issues/4417 |