Performance Test Summary

1. when testing the pay-as-you-go service, the weblogic service is disconnected when 10 users register the enterprise concurrently and run around 2 minutes 30 seconds. The reason is: the encryption and decryption API Cannot initialize the algorithm library repeatedly in the online process, solution: You have modified the API and added the unique mutex call initialization mechanism;

2. when testing the account manager, perform a reliability test on transaction log details query. After 14 hours of operation, log on to the management end using admin. When you double-click "channel type details Report ", the following message is displayed: Compilation of JSP File '/jsp/manageAccount/detailReport/channelDetail/queryChannelDetailCon. jsp 'failed:
--------------------------------------------------------------------------------
 
_ Querychanneldetailcon. java: 11: 1: cocould not write. class file for 'jsp _ servlet/_ jsp/_ manageaccount/_ detailreport/_ channeldetail/_ querychanneldetailcon ':/home/weblogic/bea/user_projects/domains/paybill /. /servers/AdminServer/tmp/_ WL_user/orgAccount/p9ep58/jsp_servlet/_ jsp/_ manageaccount/_ detailreport/_ channeldetail/_ querychanneldetailcon. class (No such file or directory)
Public final class _ querychanneldetailcon extends weblogic. servlet. jsp. jspbase implements weblogic. servlet. jsp. staleindicator {
^ Unknown ^. The cause is unknown, and no reason is found for development. Therefore, this bug is postponed.

3.

[STEP]

1. Open the tb_contract_info table in the database after performing enterprise registration and testing the performance of 10 concurrent users.
2.
3.

[Result] The information shown in the attachment diagram is displayed. The values of the two fields are "null ".

[Expected] the correct information should be displayed, not "null ".

[Note] This bug is not regular and occasionally appears.

4. Performance Testing for composite transactions

[STEP]
1. During the test.
2.
3.

[Result] The background accmgr jumps out from time to time, "scanning the database line stops abnormally and restarts...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart...
Abnormal Termination of Scanning database line, restart ..."

[Expected] This exception should not occur.

5. test the reliability of the new architecture

[STEP]
1. view the system response message in the database during the test of concurrent composite transactions.
2.
3.

[Result] The following abnormal data is displayed: "connecting to the host system service [accmgrcomm, 32001] failed", "communication fault, please go to the savings office to query", and "debit transaction: [from the savings host] the account has been paid in full. [Positive transaction: Successful] ", there are 135719 such abnormal data in 284 transactions. Although such abnormal data occurs, however, the communication with the saving host is still normal.

[Expected] The 284 abnormal data records should not exist in the database.

6. Large data volume Query

[STEP]
1. When there are 62943 transactions per day, log on to the management terminal as an admin user, query the Channel transaction details report, and query the transactions per day.
2.
3.

[Result] "operation failed: remote connection error. Please contact the Administrator", but the required data cannot be found.

[Expectation] When the data volume is large, you can query the required data for a long time.

Development processing method: when the data volume is large, the operation may occur. adjustable tuxedo and weblogic. it doesn't make much sense, because tens of thousands of pieces of data are displayed on the browser, it will let the browser down. the fundamental solution is to modify the report and change it to paging. in addition, you do not call tuxedo to directly operate on the database.

7. Send a composite transaction through performance testing

[STEP]
1. 22 hours after the reliability test is run.
2.
3.

[Result] when you make a transaction through the savings communication machine, the system prompts "communication fault. Please go to the savings office to query" and cannot make the corresponding transaction.

[Expectation] transactions should be available

Development reply: the framework issue is being processed.

8. Performance Testing for 10 concurrent users

[STEP]
1. Send money from the same city bank, with 10 users running for about 4 hours and 30 minutes.
2.
3.

[Result] An error occurred while failing to receive the background data. In this case, the database encountered a "communication fault. Please go to the savings office to query" and "connecting to the host system service [accmgrcomm, 32001] failed ", after that, the transaction can be successful when the self-help "same city bank" information is sent. At this time, the number of successful transactions in lr is 110305, and the number in the database is 110297.

[Expectation] This exception data should not occur in the database, and the number of transactions in lr should be consistent with that in the database.

Development Opinion: Because the savings communication uses the old framework, the new framework does not have this problem. However, the workload and impact of modifying the new framework are significant and will not be modified for the time being. If subsequent services cannot be met, consider replacing them.

9. Performance Test found that concurrent fatigue test errors

[STEP]
1. Use a composite script to give five users a total of 25 users for 15 hours of concurrency, and then view the background database.
2.
3.

[Result] The following error is displayed: "connecting to the host system service [accmgrcomm, 32001] failed, communication fault, please go to the savings office to query". The appendix is log and, system resource.

[Expectation] there should be no such exception information, and the cpu is sometimes idle much, which is much different from sending idle 1% at the beginning.

 

09:18:36 CPU % user % nice % system % iowait % idle
09:18:37 all 15.52 0.00 3.88 11.58 69.02
09:18:38 all 14.48 0.00 5.06 12.36 68.10
09:18:39 all 20.54 0.00 6.30 13.17 59.99
09:18:40 all 18.21 0.00 7.32 14.46 60.01

09:18:40 CPU % USER % nice % System % iowait % idle
09:18:41 all 18.68 0.00 6.12 15.30 59.90
09:18:42 all 25.03 0.00 6.57 12.89 55.51
09:18:43 all 57.49 0.00 30.09 2.31 10.11
09:18:44 all 22.84 0.00 25.16 8.57 43.43

09:18:44 CPU % user % nice % system % iowait % idle
09:18:45 all 18.94 0.00 4.62 10.25 66.19
09:18:46 all 16.02 0.00 5.38 9.51 69.09
Average: all 22.78 0.00 10.05 11.04 56.13

09:20:48 kbmemfree kbmemused % memused kbbuffers kbcached kbswpfree kbswpused % swpused kbswpcad
09:20:49 16856 8291100 99.80 1356 2249136 5526248 2666892 32.55 71828
09:20:50 16216 8291740 99.80 1320 2249952 5526248 2666892 32.55 71828
09:20:51 16024 8291932 99.81 1232 2249520 5526248 2666892 32.55 71828
09:20:52 16792 8291164 99.80 1200 2249292 5526248 2666892 32.55 71828

09:20:52 kbmemfree kbmemused % memused kbbuffers kbcached kbswpfree kbswpused % swpused kbswpcad
09:20:53 16232 8291724 99.80 1228 2249784 5526248 2666892 32.55 71828
09:20:54 15344 8292612 99.82 1288 2251284 5526248 2666892 32.55 71828
09:20:55 19376 8288580 99.77 1336 2258520 5526248 2666892 32.55 71824
09:20:56 24944 8283012 99.70 1484 2267472 5526248 2666892 32.55 71824

09:20:56 kbmemfree kbmemused % memused kbbuffers kbcached kbswpfree kbswpused % swpused kbswpcad
09:20:57 15968 8291988 99.81 1516 2255220 5526248 2666892 32.55 71824
09:20:58 15328 8292628 99.82 1544 2248432 5526248 2666892 32.55 71824
Average: 17308 8290648 99.79 1350 2252861 5526248 2666892 32.55 71826

 

Development Opinion: this bug is caused by the old program of the Framework Group. You need to consider the modification scheme in a later stage.

10. Performance Test and address query

[STEP]
1. For example, the attachment graph, recording address query script, and then use 25 users for concurrency, run for 1 hour after 34 minutes and 58 seconds.
2.
3.

[Result] The system stops when an error occurs in the attachment diagram. In this case, use ps-u front to view the service. In this case, the front service stops.

[Expected] It should not be stopped.

Development Opinion: it has something to do with the architecture of the company's framework group. Follow up on whether the framework still has this problem.

11. Performance Test compound transactions

[STEP]
1. after four and a half hours of concurrency for 25 users, the available memory is about 17 Gb/s and the loadrunner is stopped. At this time, no transaction is sent and the available memory does not increase. after about ten minutes, the available memory is about 54808K, which remains here. After the service is restarted with stopbu and startbu, the available memory is about 54k.
2.
3.

[Result]

[Expectation] When concurrent transactions are stopped, the memory used in the background should be reduced, and the available memory should be increased to the value after the service is restarted, or close to this value.

The sar command can only view the memory usage of the entire linux System and the memory usage of specific processes. In this case, use ps -- auxw to view the memory usage of the linux system. linux memory mechanism: because of the idle linux memory, it will be used by the system as a buffer, so even if you don't run any program, 99% is displayed when you use the sar command to view the memory for a long time. Therefore, this test fails.

 

12. reliable row test on the saving communication machine. The encryption mode changes from encryption to soft encryption, but cannot be traded. Cause Analysis: due to memory leakage, other programs may change the encryption mode in the memory.

Solution: since the original encryption method was to read the configuration file to the memory, subsequent transactions will directly read the configuration file in the memory, no longer read the encrypted configuration file, and then modify the program, every transaction reads the configuration file instead of storing it in the memory for read/write. The problem is solved.