Re: problems(probably) in SSL after migration from WAS3.5 to WAS5.1
On Sat, 20 Oct 2007 11:46:31 -0500, Pawel Leszczynski [EMAIL PROTECTED] wrote: Are there any other applications on this WAS instance? How are they performing? How is your JVM garbage collection? How do your RMF reports look? there are any RMF reports on WAS5.1? can you give me any sample? Assuming you are running in goal mode, and you have your tranactional CB work defined with a response-time goal, in your workload activity report your corresponding service class entry will tell you how many transactions ended during the period, etc. Tons of good info there, but what may interest you the most is the entries under waiting. The bloody infoCenter is down right now, otherwise I'd post a link with the descriptions of the waiting codes. You can also look in the Performance Monitoring and Tuning guide. There is a section on WLM Delay Monitoring. ftp://ftp.software.ibm.com/software/webserver/appserv/zos_os390/v5 1/bbo5j102.pdf If you are not familiar with how to assign your transactions to a particular transaction class/service class/reporting class, I will post a link as soon as the infoCenter is available. REPORT BY: POLICY=MYPOLICYWORKLOAD=WEBSPHERE SERVICE CLASS=WASTNX RE CRITICAL =NONE TRANSACTIONS TRANS-TIME HHH.MM.SS.TTT --DASD I/O-- ---SERVICE- --- SERVI AVG 2.61 ACTUAL 150 SSCHRT 11.0 IOC 0 CPU MPL 2.61 EXECUTION139 RESP 1.8 CPU 99003K SRB ENDED 12513 QUEUED11 CONN 0.9 MSO 0 RCT END/S 13.90 R/S AFFIN 0 DISC 0.8 SRB 0 IIT #SWAPS 0 INELIGIBLE 0 Q+PEND 0.2 TOT 99003K HST EXCTD 0 CONVERSION 0 IOSQ 0.0 /SEC 109975 AAP AVG ENC 2.61 STD DEV1.985IIP REM ENC 0.00 ABSRPTN42K MS ENC 0.00 TRX SERV 42K RESP STATE SAMPLES BREAKDOWN (% SUBP TIME --ACTIVE-- READY IDLE -WAITING TYPE(%) SUB APPL TYP3 CBBTE 0.0 0.0 0.0 0.0 0.0 0.0 CBEXE 121 5.1 89.3 0.0 0.0 5.6 GOAL: RESPONSE TIME 000.00.00.500 FOR 50% What kind of delays? how to see what kind of delays they are? What other kind of work is on this box? Is the WAS work priority reasonable with relation to your other work? Do you have the STC set to a velocity goal, and the WAS work (CB) set to a response-time goal? its definitely not a problem with CPU capacity(total utilization is very small) WAS is not waiting for CPU,I doubt also its problem with WLM settings. Its looks like a classical bottleneck but i dont know where. moreover its strange that the same application works fine on WAS3.1 and very poorly on WAS5.1 problem appears when there are many concurrent requests A lot of my questions are more background than necessarily pertinent to the question (you never know). How do you know that there are many concurrent requests? (It's much easier to track this real-time when you are using an HTTP Server front-end) When it was WAS3.5, was it also running on z/OS? What service level of WAS and Java are you running? I assume this is a webapp - no EJBs or other loveliness. So, the performance is only bad when the load is heavy? If the load then slacks off, does the performance return to acceptable levels, or does the performance just get worse as the day goes on? Do you have verbosegc turned on, and how does that output look? That should always be turned on in all of your JVMs. Are there any backend resources involved, and are they performing admirably? Do you use Tivoli Performance Viewer, or whatever it was called in 5.1? If you'd rather respond off-list, that's fine. Aaron -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: problems(probably) in SSL after migration from WAS3.5 to WAS5.1
hi Aaron, I also found this: http://www-1.ibm.com/support/docview.wss?uid=swg21215703 Are you running through the z/OS HTTP Server? no, we are using WAS5.1 without front-ending with HTTP server If so, is the communication between that and WAS using SSL (if you don't know, then it probably is). That can be changed to HTTP if you want - save you some cycles. Do you have crypto engines configured and used? Any other specialty engines? we have crypto hardware and RMF reports shows that its utilized Are there any other applications on this WAS instance? How are they performing? How is your JVM garbage collection? How do your RMF reports look? there are any RMF reports on WAS5.1? can you give me any sample? What kind of delays? how to see what kind of delays they are? What other kind of work is on this box? Is the WAS work priority reasonable with relation to your other work? Do you have the STC set to a velocity goal, and the WAS work (CB) set to a response-time goal? its definitely not a problem with CPU capacity(total utilization is very small) WAS is not waiting for CPU,I doubt also its problem with WLM settings. Its looks like a classical bottleneck but i dont know where. moreover its strange that the same application works fine on WAS3.1 and very poorly on WAS5.1 problem appears when there are many concurrent requests thanks, Pawel Aaron -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
problems(probably) in SSL after migration from WAS3.5 to WAS5.1
Hello everybody, After migration one of our application from WAS3.5 to WAS5.1 we encountered following problem: response time grows dramatically, and in time of peak, most of users cannot simply logon (or work with) application. In WAS log we can see a lot of messages like that: Trace: 2007/10/18 10:47:51.624 01 t=AC40D0 c=UNK key=S2 (0E012048) Description: Failure Exit from - SecurityManager::secureSocketInit Socket descriptor: 525 gsk_secure_socket_init: 415 we found on IBM's pages solution of similar problem: http://www-1.ibm.com/support/docview.wss?uid=swg1PK34416 but 1. we don't use PAUSELISTENER nor RESUMELISTENER modify commands 2. environment variable control_region_dreg_on_no_srs is not specified (so we don't know if it's really this situation). One more question: we looked at our configuration of WAS5.1 and some doubts appear: 1.control_region_ssl_thread_pool_size is set to default = 3 (in documentation is written that it shouldn't be changed) (isn't it too small for big workload?) 2.control_region_thread_pool_size=25. (maybe here is bottleneck?) thanks for any suggestions, Pawel Leszczynski PKO BP SA -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: problems(probably) in SSL after migration from WAS3.5 to WAS5.1
I also found this: http://www-1.ibm.com/support/docview.wss?uid=swg21215703 Are you running through the z/OS HTTP Server? If so, is the communication between that and WAS using SSL (if you don't know, then it probably is). That can be changed to HTTP if you want - save you some cycles. Do you have crypto engines configured and used? Any other specialty engines? Are there any other applications on this WAS instance? How are they performing? How is your JVM garbage collection? How do your RMF reports look? What kind of delays? What other kind of work is on this box? Is the WAS work priority reasonable with relation to your other work? Do you have the STC set to a velocity goal, and the WAS work (CB) set to a response-time goal? Aaron On Fri, 19 Oct 2007 07:38:23 -0500, Pawel Leszczynski [EMAIL PROTECTED] wrote: Hello everybody, After migration one of our application from WAS3.5 to WAS5.1 we encountered following problem: response time grows dramatically, and in time of peak, most of users cannot simply logon (or work with) application. In WAS log we can see a lot of messages like that: Trace: 2007/10/18 10:47:51.624 01 t=AC40D0 c=UNK key=S2 (0E012048) Description: Failure Exit from - SecurityManager::secureSocketInit Socket descriptor: 525 gsk_secure_socket_init: 415 we found on IBM's pages solution of similar problem: http://www-1.ibm.com/support/docview.wss?uid=swg1PK34416 but 1. we don't use PAUSELISTENER nor RESUMELISTENER modify commands 2. environment variable control_region_dreg_on_no_srs is not specified (so we don't know if it's really this situation). One more question: we looked at our configuration of WAS5.1 and some doubts appear: 1.control_region_ssl_thread_pool_size is set to default = 3 (in documentation is written that it shouldn't be changed) (isn't it too small for big workload?) 2.control_region_thread_pool_size=25. (maybe here is bottleneck?) -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html