-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: 7.0.0
-
Fix Version/s: 7.0.0
-
Component/s: module - Notifications
-
Labels:
-
Environment:IDM runs on a CentOS7 with Java 11, external DJ as repo.
Noticed about 6% throughput drop and 7% mean response times drop in creating simple managed users using PUT test introduced by the commit 015ba931:
Before the commit 015ba931, Gatling tool has the following report for the test:
---- Global Information -------------------------------------------------------- > request count 163199 (OK=163199 KO=0 ) > min response time 27 (OK=27 KO=- ) > max response time 555 (OK=555 KO=- ) > mean response time 55 (OK=55 KO=- ) > std deviation 13 (OK=13 KO=- ) > response time 50th percentile 53 (OK=53 KO=- ) > response time 95th percentile 74 (OK=74 KO=- ) > response time 99th percentile 90 (OK=90 KO=- ) > response time 99.9th percentile 138 (OK=138 KO=- ) > mean requests/sec 542.189 (OK=542.189 KO=- ) ---- Response Time Distribution ------------------------------------------------ > t < 800 ms 163199 (100%) > 800 ms < t < 1200 ms 0 ( 0%) > t > 1200 ms 0 ( 0%) > failed 0 ( 0%)
With commit 015ba931: We have
---- Global Information -------------------------------------------------------- > request count 152746 (OK=152746 KO=0 ) > min response time 30 (OK=30 KO=- ) > max response time 559 (OK=559 KO=- ) > mean response time 59 (OK=59 KO=- ) > std deviation 13 (OK=13 KO=- ) > response time 50th percentile 57 (OK=57 KO=- ) > response time 95th percentile 78 (OK=78 KO=- ) > response time 99th percentile 97 (OK=97 KO=- ) > response time 99.9th percentile 157 (OK=157 KO=- ) > mean requests/sec 507.462 (OK=507.462 KO=- ) ---- Response Time Distribution ------------------------------------------------ > t < 800 ms 152746 (100%) > 800 ms < t < 1200 ms 0 ( 0%) > t > 1200 ms 0 ( 0%) > failed 0 ( 0%)
Note: The test used the external DJ(latest DJ master build) as repo, if the changes are not repo dependent, the issue should be applicable to JDBC repo as well(see degradation for JDBC repo as well in Jenkins test job runs)
To reproduce the symptom using Pyforge:
1. Set Pyforge configurations as below:
[Stress]
num_users = 10
duration = 60
concurrency = 10
max_throughput = -1
test_duration = 300
preload_users = 20
preload_duration = 300
test_concurrency = 30
[OpenIDM]
host_name = ${Default:host_name}
java_home = /usr/java/jdk11
java_args = -Xms6g -Xmx6g
version = 7.0.0-SNAPSHOT
previous_version = 6.5.0
protocol = http
repo_type = external_dj
use_docker_repo = False
jvm_debug_options = ${Default:jvm_debug_options}
cloud = ${Default:cloud}
enable_metrics = True
enable_policy_enforcement = False
enable_jetty_config = False
[IDMExternalDJ]
host_name = ${Default:host_name}
java_home = ${Default:java_home}
java_args = ${Default:java_args}
version = 7.0.0-SNAPSHOT
ldap_port = 31389
base_dn = "dc=openidm,dc=forgerock,dc=com"
admin_port = 34444
2. run the test as:
run-pybot.py -v -c perf -s simple_managed_user.CreateManUsersPut OpenIDM
3. Check results in pyforge/latest/debug.txt
Attached are the metrics for the two("good"/"bad") commits.
- is caused by
-
-
- Closed
-
- is duplicated by
-
-
- Closed
-
-
-
- Closed
-
- is related to
-
-
- Closed
-