William, thanks for sharing this - it will probably help others at some time.
Re: Lesson Three. Could you elaborate on the apps "helping" you by changing the DB CLOB settings? This is disturbing. Rick On Fri, Aug 15, 2008 at 8:28 AM, William Rentfrow < [EMAIL PROTECTED]> wrote: > ** This one was a proverbial hum-dinger to fix. It's an interesting > problem though so I'll go back through it. > > Server: ARS 7.1 on Solaris with Incident Management 7.03, SLM 7.1, normal > ancillary apps such as AE, EE, MT, etc > 4 MT servers on Solaris with Websphere/IBM HTTP server > > The original problem: Users were experiencing a "Unique index violation" > when saving an incident. The only failure noted in any of the logs client > or server side was an API -CE entry that said "Fail". > > *Lesson one:* The one I have to re-learn every couple of years - always > ask your users "Were there any other error messages you received?" > > Users will always tell you the LAST error message they got and not the > first necessarily. In this case the answer was "Yes - I got an RPC timeout > message first". > > I didn't get that piece of information until day 3 of this problem. > > *Lesson two:* Sometimes an error is correct. > > In my case the unique index error WAS correct. The users were hitting > "Save" and then getting an RPC timeout after 2 minutes or so of not > receiving a response from the AR Server. In the meantime the AR Server was > receiving the Incident and saving it. The user - who had only seen an RPC > timeout message - was still in the "new" mode of the system and hit save > again which correctly generated the unique index violation message. The > instanceID (179) would cause this error - so would the Incident Number. > Both were duplicated in the second save attempt and Oracle said "NO." > > *Lesson three:* If you are using Oracle remotely with Remedy it is > imperative you switch to the in-row CLOB storage option. > > Once I figured #2 above out I had to deal with the performance issue. We > had previously had major performance problems on our QA server. The > symptoms were a <long string of expletives goes here> to deal with - all you > saw in any of the log files was a huge gap while Oracle was doing > transactions. Literally nothing was going on. > > When we built our PRD system we switched the in-row option immediately > after installing the AR Server - then we installed the apps. The apps were > nice enough to "correct" our setting back to out-of-row and we didn't notice > until this happened. We ended up doing an emergency re-structuring of the > database once I was able to show again in the logs there were long, long > gaps (like 10 seconds total per transaction) in the Submit/Modify processes. > > *Lesson four:* All that improved the problem but still didn't fix it. > After much more log reading and noticing very small gaps ( < 50 ms) between > API transactions I upped the fast/list queues by 25% to 16 each. We run > 3000 incidents/day through this system with several hundred users. > > Voila. Problem fixed. > > Must............enjoy..................weekend..................... > > William Rentfrow, Principal Consultant > [EMAIL PROTECTED] > C 701-306-6157 > O 952-432-0227 > > > __Platinum Sponsor: www.rmsportal.com ARSlist: "Where the Answers Are" > html___ _______________________________________________________________________________ UNSUBSCRIBE or access ARSlist Archives at www.arslist.org Platinum Sponsor: www.rmsportal.com ARSlist: "Where the Answers Are"