Re: A teensy-weensy little oversight in the QA suite?

On 29 Jan, 14:01, Fernando Nunes <domusonl...@gmail.com> wrote: > Accordingly to the report, if you need to upgrade TODAY, then use xC6 and > use the workaround (force the building of the SMI). > If you upgrade after xC6W1 is available then use it and you won't need the > workaround. > > Regards. > > On Fri, Jan 29, 2010 at 10:28 AM, Bartlomiej Lidke > > > > <ohggu...@rcs.cy.rot13.invalid> wrote: > > Obnoxio The Clown <obno...@serendipita.com> wrote: > > >http://www-01.ibm.com/support/docview.wss?uid=swg21418229&myns=swgimg... > > > > /me bangs head on desk > > > try to find a 'recommended' word here: > > >http://www-01.ibm.com/support/docview.wss?uid=swg27014361#11.50 > > > 11.50.xC5W4     = Superseded > > 11.50.xC6       = Will be replaced by 11.50.xC6W1 > > 11.50.xC6W1     = Future release > > > :-( > > > -- > > butthead > > _______________________________________________ > > Informix-list mailing list > > Informix-l...@iiug.org > >http://www.iiug.org/mailman/listinfo/informix-list > > -- > Fernando Nunes > Portugal > > http://informix-technology.blogspot.com > My email works... but I don't check it frequently... Well even that is not ok.. IC64331 AFTER CONVERSION TO 11.50 THE ONCHECK -CC COMMAND REPORTS A DISCREPANCY IN NEXT EXTENT SIZE FOR SOME SYSTEM CATALOG TABL IC63942 IN-PLACE MIGRATION WITH AUDITING ACTIVE CAN LEAD TO ASSERTION FAILURE: PTALTSTAT2 BAD ALTER SLOT IC64745 WHEN MIGRATING FROM 7.31.FD6 TO 11.50.FC5 WITH DATABASES PAGE ZERO IS OVERWRITTEN WITH PAGE 4 OR 5 IC64185 MIGRATION OF A FUNCTION RETURNS 211 ERROR ASSERT FAILS AND CREATES CORRUPTION IN THE SYSTEM CATALOG IC60800 OUTPUT FILE CONTAINING ERRORS FROM VIOLATION TABLE CONVERSION IS OVERWRITTEN WHEN A SUBSEQUENT DATABASE IS CONVERTED So upgradean can corrupt the system catalog three ways and cause some errors to be missed. IC63876 APPARENT RACE CONDITION WHEN BAR_MAX_BACKUP WORKER PROCESSES RETURN PROCESSES OUT OF SEQUENCE IC64918 ONTAPE THREAD CAN HANG IN PH_SUBMIT_TASK AFTER COMPLETING LOG BACKUP So version 11 allows parallel backups even on unlogged databases but they do not work reliably, ontape log backups are not reliable IC60893 IPLOAD CORE DUMPS ON 11.50 There are no details but HPL is not reliable IC62958 CERTAIN ONAUDIT ERRORS CAN CAUSE RAPID INCREASE IN THE ONLINE LOG SIZE HIGH MEMORY CONSUMPTION AND EXTENSIVE ALARMPROGRAM CALL IC61295 HUGE MEMORY PEAKS ON SYSTEM WITH LARGE NUMBER OF TABLES/ FRAGMENTS DESPITE MOST AGGRESSIVE PTCLOSER CONFIGURATION IC62299 SEQUENCE OBJECT USE IN A NON-LOGGING DATABASE CAN FILL LOGS WITH UNIQ8ID AND CAUSE LONG TRANSACTIONS WHEN USING NOCACHE OPTION Three issues can cause resource overallocation IC62244 ADDING LOGICAL LOGS AFTER A TRANSACTION REACHES LTXEHWM DOES NOT RELEASE THE SERVER FROM BLOCKED:LONGTX STATE Logging can get stuck and is therefore not reliable. IC64811 WHEN USING SQLTRACE SQL_RUNTIME SQL_TOTALTIME AND OTHER TIME FIELDS ARE INCORRECT SQL TRACE does not work IC64908 ON STARTUP OF SERVER ON LINUX ONINIT PRINTS ERROR PID CANT GET REAL PATH OF Startup gives errors on Linux IC63252 A STORED PROCEDURE CAN TAKE TOO LONG TO COMPLETE WHEN CALLING NESTED PROCEDURES Nested stored procs can be slow. IC64421 AFTER RAISE EXCEPTION IN TRIGGER PROCEDURE A CONSTRAINT VIOLATION IS APPLIED TO THE NEXT STATEMENT Triggers with constraints do not work IC64805 INSERT ROW FAILED IN MKROWID SQLERR RECEIVED -626 Inserts do not work!! IC64937 WHEN ASSIGNING A STRING OF INCORRECT SIZE TO A VARCHAR VARIABLE NO ERROR IS RETURNED AND CAN LEAD TO MEMORY OVERFLOW Memory overflows can occur and are not be trapped (I assume as with all memory overflows this can cause server srashes) IC62826 OPTIMIZER INCORRECTLY PERFORMS SEQUENTIAL SCAN ON TABLE RESULTING IN SLOW QUERY TIME Optimizer is broken http://www-01.ibm.com/support/docview.wss?rs=0&uid=swg27014361 indicates that the next proper fully tested release will be 11.50.xC7 estimated for Apr 12 and since fixpacks always seem to slip a month then that will be beginning of May. So 4 months to deploy, do all testing and upgrade before version 10 support ends..not good. IBM NEED TO FIX ALL THE ABOVE ISSUES in 11.50.xC7 WITHOUT FAIL.

<david@smooth1.co.uk> wrote in message news:1756ee51-b3d3-4adf-97e9-6bf683c590b3@b10g2000yqa.googlegroups.com... On 29 Jan, 14:01, Fernando Nunes <domusonl...@gmail.com> wrote: > Accordingly to the report, if you need to upgrade TODAY, then use xC6 and > use the workaround (force the building of the SMI). > If you upgrade after xC6W1 is available then use it and you won't need the > workaround. >> Well even that is not ok.. <list of bugs> >> IBM NEED TO FIX ALL THE ABOVE ISSUES in 11.50.xC7 WITHOUT FAIL. Or else, you will do what? In v10, we are now quite stable ... in FC10! Here is one of the drawbacks of low useage of IDS; it takes a long time to learn lesssons that will be found only "in the field". This, I expect, is even worse in the "premium" products such as MACH11, which surely get very little use in "real-world" circumstances?

david@smooth1.co.uk wrote: > On 29 Jan, 14:01, Fernando Nunes <domusonl...@gmail.com> wrote: >> Accordingly to the report, if you need to upgrade TODAY, then use xC6 and >> use the workaround (force the building of the SMI). >> If you upgrade after xC6W1 is available then use it and you won't need the >> workaround. >> >> Regards. >> >> On Fri, Jan 29, 2010 at 10:28 AM, Bartlomiej Lidke >> >> >> >> <ohggu...@rcs.cy.rot13.invalid> wrote: >>> Obnoxio The Clown <obno...@serendipita.com> wrote: >>> http://www-01.ibm.com/support/docview.wss?uid=swg21418229&myns=swgimg... >>>> /me bangs head on desk >>> try to find a 'recommended' word here: >>> http://www-01.ibm.com/support/docview.wss?uid=swg27014361#11.50 >>> 11.50.xC5W4 = Superseded >>> 11.50.xC6 = Will be replaced by 11.50.xC6W1 >>> 11.50.xC6W1 = Future release >>> :-( >>> -- >>> butthead >>> _______________________________________________ >>> Informix-list mailing list >>> Informix-l...@iiug.org >>> http://www.iiug.org/mailman/listinfo/informix-list >> -- >> Fernando Nunes >> Portugal >> >> http://informix-technology.blogspot.com >> My email works... but I don't check it frequently... > > Well even that is not ok.. > > > IC64331 AFTER CONVERSION TO 11.50 THE ONCHECK -CC COMMAND REPORTS A > DISCREPANCY IN NEXT EXTENT SIZE FOR SOME SYSTEM CATALOG TABL This will not happen in every system. And it's a discrepancy between information in the systables and the real table layout > > IC63942 IN-PLACE MIGRATION WITH AUDITING ACTIVE CAN LEAD TO ASSERTION > FAILURE: > PTALTSTAT2 BAD ALTER SLOT This requires a pre-version 9.x to be in-place migrated to v9.x and then to v10 or v11.x. It will depend on the audit level used, and it will depend on "bad luck" (thread running sequence, because the "issue" will be fixed during conversion). So it's not even v11 specific... > > IC64745 WHEN MIGRATING FROM 7.31.FD6 TO 11.50.FC5 WITH DATABASES PAGE > ZERO IS OVERWRITTEN > WITH PAGE 4 OR 5 Seems to be platform specific, but I can't be sure. In any case, it's nice to see some work done related to non-supported versions (7.31). > > IC64185 MIGRATION OF A FUNCTION RETURNS 211 ERROR ASSERT FAILS AND > CREATES > CORRUPTION IN THE SYSTEM CATALOG Happens converting v10 to v11.. > > IC60800 OUTPUT FILE CONTAINING ERRORS FROM VIOLATION TABLE CONVERSION > IS > OVERWRITTEN WHEN A SUBSEQUENT DATABASE IS CONVERTED > Errors occurred during convertion of violating tables. These errors will not happen most of the times. > So upgradean can corrupt the system catalog three ways and cause some > errors to be missed. > > IC63876 APPARENT RACE CONDITION WHEN BAR_MAX_BACKUP WORKER PROCESSES > RETURN PROCESSES OUT OF SEQUENCE The PMR was closed after some OS patches were installed. The problem didn't happen again. > > IC64918 ONTAPE THREAD CAN HANG IN PH_SUBMIT_TASK AFTER COMPLETING LOG > BACKUP It CAN hang. Needs some specific thread schedule timmings. A debugger was needed to rreproduce this... > > So version 11 allows parallel backups even on unlogged databases but > they do not work reliably, ontape log backups are not reliable > > IC60893 IPLOAD CORE DUMPS ON 11.50 It will depend on the x-Server you use. I hit this one. It has a pretty easy workaround (don't use 32bit color), but it can be a PITA to figure what is happening. Several x-servers can handle the issues, others don't. I'm happy to see we can fix it in the engine. > > There are no details but HPL is not reliable > > IC62958 CERTAIN ONAUDIT ERRORS CAN CAUSE RAPID INCREASE IN THE ONLINE > LOG SIZE HIGH > MEMORY CONSUMPTION AND EXTENSIVE ALARMPROGRAM CALL I like this one. The interesting part of this is that the real issue has nothing to to with onaudit... It should turn out to be an alarm sub-system feature (or maybe bug). Some changes in the alarmprogram could ease the situation but not solve it completely. The fact that it can happen with other situations, makes me wonder how many times it happened to each of us? Oh... and it's not 11.50 specific. > > IC61295 HUGE MEMORY PEAKS ON SYSTEM WITH LARGE NUMBER OF TABLES/ > FRAGMENTS DESPITE MOST > AGGRESSIVE PTCLOSER CONFIGURATION This is not a problem that causes resource overallocation. This overallocation (assuming we accept its more than it should be) has seen some features to decrease them. In some environments these features apparently are not enough. Maybe the customer needs more hardware to handle his database load. But R&D will keep trying... Notice that the problem is not new. The solution is relatively new, and apparently in some situation it's still not enough. > > IC62299 SEQUENCE OBJECT USE IN A NON-LOGGING DATABASE CAN FILL LOGS > WITH UNIQ8ID > AND CAUSE LONG TRANSACTIONS WHEN USING NOCACHE OPTION This turned out to be expected behavior. In non-logged databases, some activities must be logged... Sequences are one of them. In this particular situation, the customer was doing an insert of 93M rows. Each row will require a logical log entry. Without the NOCACHE, one entry will take 20 sequence "steps". So without NOCACHE the logical log activity was much (20x) higher. > > Three issues can cause resource overallocation > > IC62244 ADDING LOGICAL LOGS AFTER A TRANSACTION REACHES LTXEHWM DOES > NOT RELEASE THE > SERVER FROM BLOCKED:LONGTX STATE This is "normal" behavior. The defect associated with the APAR is classified as "feature". The behavior described is there since DYNAMIC_LOGS was introduced. Now... If you ask me, I'd give my strong vote for this feature. But it's nothing specific for v11 (except that a possible fix/feature will only appear in some 11.x+ of course) > > Logging can get stuck and is therefore not reliable. Logging has always been "stuckable" if you don't configure it properly. No news here. > > IC64811 WHEN USING SQLTRACE SQL_RUNTIME SQL_TOTALTIME AND OTHER TIME > FIELDS ARE INCORRECT Hmmm... I've also seen this one. It was during test/benchmark phase of a 7.31 to 11.50 migration. It was inconvenient, but didn't stop me from using SQLTRACE to show the customer what was the issue. They couldn't solve it in the application (where the issue was generated), but some engine tweaking was enough to overcome their doubts. Running happily on 11.50 for some months. > > SQL TRACE does not work ... at 100% > > IC64908 ON STARTUP OF SERVER ON LINUX ONINIT PRINTS ERROR PID CANT GET > REAL PATH OF > In practice it's a cosmetic issue. It doesn't happen in every Linux box. It's also the tip of an Iceberg (not related to IDS): http://blogs.igalia.com/aperez/2009/01/the-story-of-linux-gatevdsoso/ IDS apparently is trying to call stat() on it... > Startup gives errors on Linux ... on some systems, and then apparently works as expected. Not sure why we do the stat(), and what happens when we can't do it... > > > IC6325

De Stomerij wrote: > <david@smooth1.co.uk> wrote in message > news:1756ee51-b3d3-4adf-97e9-6bf683c590b3@b10g2000yqa.googlegroups.com... > On 29 Jan, 14:01, Fernando Nunes <domusonl...@gmail.com> wrote: >> Accordingly to the report, if you need to upgrade TODAY, then use xC6 and >> use the workaround (force the building of the SMI). >> If you upgrade after xC6W1 is available then use it and you won't need >> the >> workaround. > >>> Well even that is not ok.. > <list of bugs> >>> IBM NEED TO FIX ALL THE ABOVE ISSUES in 11.50.xC7 WITHOUT FAIL. > > Or else, you will do what? > > In v10, we are now quite stable ... in FC10! > > Here is one of the drawbacks of low useage of IDS; it takes a long time > to learn lesssons that will be found only "in the field". > > This, I expect, is even worse in the "premium" products such as MACH11, > which surely get very little use in "real-world" circumstances? > Complementing my answer to David: - I have customer running "quite stable" with 11.50 (FC5). - Regarding your observation about MACH 11, it depends. HDR and RSS are VERY common. ER is not so common. Connection Manager is becoming comming in anyone using HDR/RSS. This discussion remembers me of an epic discussion I, some people from a customer, and a consultant from a competitor had. This consultant was saying that a patch simply corrects a bug and never introduces other issues. I tried to reason with him, and I was followed by the customer team. No luck... He kept is view. Needless to say that since then, both me, and (specially) the customer had several situations where we reminded him of his now famous position... ;) Now, if a patch can bring nasty errors, of course new code can do the same. And, although we call it software "engineering", this is very far from the certainty we get in buildings, ships, planes, roads etc. (other forms of engineering) Regards.

"Fernando Nunes" <domusonline@gmail.com> wrote in message news:hk5i48$9ef$1@news.eternal-september.org... >> This, I expect, is even worse in the "premium" products such as MACH11, >> which surely get very little use in "real-world" circumstances? >> > > Complementing my answer to David: > > - I have customer running "quite stable" with 11.50 (FC5). > - Regarding your observation about MACH 11, it depends. HDR and RSS are > VERY common. ER is not so common. Connection Manager is becoming comming > in anyone using HDR/RSS. Yes, I should have not mentioned MACH11. Instead, I mean data compression and "continuous access feature" (CAF/"RAC"). Both these are, I believe, chargeable even on top of the Enterprise fee. Is the uptake high? I know of no-one using them.