Jmeter - 1000 concurrent users are not created at the same time

Question

I wrote a jmeter script, where thread count is 1000, ramp-up time is 1, loop is forever, duration is an hour.

When I run this script I see that jmeter starts creating threads, and as I set loop to forever, so non of the threads should be completed before the end of duration.

But the problem is that jmeter cannot create those threads concurrently. I added short log details from summary.

Totally 76 threads were created in 1 minute. Totally 264 threads were created in 10 minutes. Totally 590 threads were created in 59 minutes.

Then it started to end the existing threads as it already was an hour, and next started to create other 510 threads. And at the end, it showed that 1000 threads were created.

But actually it had to create them concurrently and maybe it could take just 3-5 minutes, otherwise it had to fail.

I do these test in AWS with m3.medium instance.

Before that, I used t2.micro server with 100 threads in an hour. And everything was fine, I mean all threads were created in few minutes and were active till within hour.

I added the last lines from summary.

summary +    840 in 00:00:25 =   33.1/s Avg:  6989 Min:    19 Max: 69102 Err:     3 (0.36%) Active: 591 Started: 591 Finished: 0
summary = 424179 in 00:59:03 =  119.7/s Avg:  2963 Min:    19 Max: 137203 Err:   498 (0.12%)
summary +   1542 in 00:01:31 =   17.0/s Avg: 36302 Min:    19 Max: 136434 Err:   388 (25.16%) Active: 4 Started: 683 Finished: 679
summary = 425721 in 01:00:33 =  117.2/s Avg:  3084 Min:    19 Max: 137203 Err:   886 (0.21%)
summary +    171 in 00:00:30 =    5.7/s Avg:   367 Min:   178 Max:   920 Err:     0 (0.00%) Active: 8 Started: 858 Finished: 850
summary = 425892 in 01:01:03 =  116.3/s Avg:  3082 Min:    19 Max: 137203 Err:   886 (0.21%)
summary +    149 in 00:00:24 =    6.1/s Avg:   352 Min:   162 Max:   992 Err:     0 (0.00%) Active: 0 Started: 1000 Finished: 1000
summary = 426041 in 01:01:28 =  115.5/s Avg:  3081 Min:    19 Max: 137203 Err:   886 (0.21%)

So could anyone suggest the solution for this situation, please?

Thread Group - screenshot

summary +   1507 in 00:01:30 =   16.7/s Avg: 34686 Min:    23 Max: 116663 Err:   258 (17.12%) Active: 588 Started: 588 Finished: 0
summary = 404415 in 00:57:31 =  117.2/s Avg:  3111 Min:    19 Max: 116663 Err:   674 (0.17%)
summary +    378 in 00:00:27 =   14.0/s Avg: 31968 Min:    33 Max: 112930 Err:    54 (14.29%) Active: 589 Started: 589 Finished: 0
summary = 404793 in 00:57:58 =  116.4/s Avg:  3138 Min:    19 Max: 116663 Err:   728 (0.18%)
summary +   1868 in 00:01:30 =   20.7/s Avg: 30655 Min:    19 Max: 116585 Err:   126 (6.75%) Active: 592 Started: 592 Finished: 0
summary = 406661 in 00:59:28 =  114.0/s Avg:  3265 Min:    19 Max: 116663 Err:   854 (0.21%)
summary +    896 in 00:01:06 =   13.7/s Avg: 21775 Min:    19 Max: 90675 Err:     8 (0.89%) Active: 476 Started: 593 Finished: 117
summary = 407557 in 01:00:33 =  112.2/s Avg:  3305 Min:    19 Max: 116663 Err:   862 (0.21%)
summary +    537 in 00:00:24 =   22.2/s Avg: 30182 Min:   171 Max: 88596 Err:    85 (15.83%) Active: 2 Started: 701 Finished: 699
summary = 408094 in 01:00:58 =  111.6/s Avg:  3341 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    165 in 00:00:30 =    5.5/s Avg:   288 Min:   167 Max:   803 Err:     0 (0.00%) Active: 4 Started: 868 Finished: 864
...
...
summary = 412193 in 01:12:28 =   94.8/s Avg:  3311 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    178 in 00:00:30 =    5.9/s Avg:   334 Min:   169 Max:   959 Err:     0 (0.00%) Active: 3 Started: 4979 Finished: 4976
summary = 412371 in 01:12:58 =   94.2/s Avg:  3309 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    178 in 00:00:30 =    6.0/s Avg:   326 Min:   172 Max:  1072 Err:     0 (0.00%) Active: 2 Started: 5156 Finished: 5154
summary = 412549 in 01:13:28 =   93.6/s Avg:  3308 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    175 in 00:00:30 =    5.8/s Avg:   343 Min:   173 Max:  1066 Err:     0 (0.00%) Active: 3 Started: 5332 Finished: 5329
summary = 412724 in 01:13:58 =   93.0/s Avg:  3307 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    168 in 00:00:30 =    5.6/s Avg:   352 Min:   175 Max:  1057 Err:     0 (0.00%) Active: 4 Started: 5501 Finished: 5497
summary = 412892 in 01:14:28 =   92.4/s Avg:  3306 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    179 in 00:00:30 =    6.0/s Avg:   335 Min:   168 Max:   914 Err:     0 (0.00%) Active: 1 Started: 5677 Finished: 5676
summary = 413071 in 01:14:58 =   91.8/s Avg:  3304 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    178 in 00:00:30 =    5.9/s Avg:   306 Min:   169 Max:   984 Err:     0 (0.00%) Active: 2 Started: 5856 Finished: 5854
...
...
summary = 416548 in 01:24:58 =   81.7/s Avg:  3279 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    175 in 00:00:30 =    5.9/s Avg:   322 Min:   172 Max:   935 Err:     0 (0.00%) Active: 3 Started: 9331 Finished: 9328
summary = 416723 in 01:25:28 =   81.3/s Avg:  3278 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    172 in 00:00:30 =    5.7/s Avg:   320 Min:   163 Max:   902 Err:     0 (0.00%) Active: 6 Started: 9506 Finished: 9500
summary = 416895 in 01:25:58 =   80.8/s Avg:  3277 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    175 in 00:00:30 =    5.8/s Avg:   319 Min:   168 Max:   908 Err:     0 (0.00%) Active: 3 Started: 9678 Finished: 9675
summary = 417070 in 01:26:28 =   80.4/s Avg:  3276 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    170 in 00:00:30 =    5.7/s Avg:   321 Min:   165 Max:  1287 Err:     0 (0.00%) Active: 4 Started: 9849 Finished: 9845
summary = 417240 in 01:26:58 =   80.0/s Avg:  3275 Min:    19 Max: 116663 Err:   947 (0.23%)
summary +    154 in 00:00:26 =    5.9/s Avg:   325 Min:   165 Max:   979 Err:     0 (0.00%) Active: 0 Started: 10000 Finished: 10000
summary = 417394 in 01:27:24 =   79.6/s Avg:  3273 Min:    19 Max: 116663 Err:   947 (0.23%)
Tidying up ...    @ Sun Jul 30 09:16:26 UTC 2017 (1501406186446)
... end of run

definitely starting 1000 threads in 1 second is an uphill task for any computer, some even lead to a crash and it entirely depends on system resources like CPU, memory etc. it is practically not possible to simulate unless u have a super computer. please share the Thread Group screenshot. — Naveen Kumar R B

Dmitri T Dmitri T · Accepted Answer · 2017-08-01T11:44:42

Loooking into Amazon instances specifications:

Model   vCPU    Mem (GiB)   SSD Storage (GB)

m3.medium   1   3.75    1 x 4

I strongly doubt you will be able to kick off 1000 threads in 1 second as most likely your instance will run out of resources, double check its health metrics using i.e. Amazon CloudWatch or JMeter's PerfMon Plugin

So make sure that your JMeter load generator has enough spare resources as if JMeter instance will be overloaded it won't be able to send requests fast enough so you will see very large response times although your application under test may behave fine.

I would recommend adding another instance and running JMeter in distributed mode, this way you should be in a more safe position.

Jmeter - 1000 concurrent users are not created at the same time

1 Answers