Want some flame?
Wolf Duttlinger-Manger
thanks for your comments. PMR is right. I do totally agree - do not
bother development with everyday problems.
In our case we spend two days to find out what was wrong. I tried any
combination of restore/rollforward I can think of.
I don't even say I need to talk to a developer. What I needed was:
- analyse what is wrong with log 607 - syntax check the contents,
possibily dump the contents
- correct the problem with rollforward
- or, get the online backup out of rollforward state.
What did I expect?
Check db2diag. Analyse the error messages. Know (!) what happend. Find a
way to fix it.
Instead I had to "restore to the same date (16.th, 11:59pm), restore to
2h before crash, restore to past year 2160....". They questioned whether
my installation was ok because error messages were in mixed language. I
had to create a dummy db and copy SQLOGCTL.LFH over.
In the end I had to go through discussions as "but there were several
crashes before....". Either a DB is designed to recover after a crash,
or it is not.
What happend with db2reset is ridiculous!
To me this is a problem where either development has to supply tools or
has to take over - immediatelly.
I actually did ask for the file layout of the SQLOGCTL.LFH and the LOG
files - I would have tried to "db2reset" the DB myself. I would have
been on my own - but at least it would have been possible to "rescue"
some data.
Can you imagine - I know the data is there, I can "smell" it, feel it -
but have no access to it.
And so we lost four days of work and several orders!!!
Conclusion - this summer we were to evaluate the migration from Oracle
to DB2 in a SAP environment - I can't see anybody in our IT wishing to
ask for resources on this evaluation....
Grüße aus dem Schwarzwald.
Wolf
Serge Rielau wrote:
>
> Hallo Wolf,
>
> I took the liberty and followed up here in Toronto.
> Assuming you talk of PMR 68747,070 this one
> reached Toronto Lab. and it is marked as being worked on.
> India is responsible for the European market providing
> the same service that service in Toronto does provide.
> PMRs get cascaded to service in Toronto during off hours in India.
> If a PMR cannot get contained by service it cascades to us developers
> who may be in Toronto, San Jose, Boeblingen, or elsewhere.
> Before service has assessed exactly what is going wrong and that they can't
> fix it
> is not helpful to contact development because service has the big picture
> and
> the expertise with the system. Furthermore they have to pick the right
> developers for the specific problem.
>
> Hopefully your experience turns out to be the big exception and I'm certain
> service
> will forward the conclusions to development, to further decrease likelihood
> of such an event in the future.
>
> Gruss nach D'Land von einem Horber
> Serge
>
> Wolf Duttlinger-Manger wrote:
>
> > I won't ever recommend using DB2 again!
> >
> > Here's what happend to me - feel free to comment on this:
> >
> > Thursday, 17th - 10:30 am - System crash
> > System rebooted
> > DB2 RESTART DATABASE -> Crash recovery needed -> finishes with DIA3910W
> > and rc -6899 - whatever this means
> > Result - database being marked as bad
> >
> > Well - I DO have a set off backup's. Online!
> >
> > db2 RESTORE DATABASE AWT USE ADSM OPEN 1 SESSIONS TAKEN
> > AT 20000216220001 WITH 2 BUFFERS BUFFER 1024 WITHOUT PROMPTING
> > ->ok
> > watch out now:
> > db2 rollforward database awt to end of logs and stop
> > SQL4970N Roll-forward recovery on database "AWT" cannot reach
> > the specified stop point (end-of-log or point-in-time)
> > because of missing log file(s) on node(s)
> > "0".
> >
> > Great!
> >
> > Obviously somethings is wrong with the last log-file.
> >
> > From now on I did spend two days with IBM support trying to access the
> > data that is there:
> > Online Backup as of 16th 10pm - nobody did work at that time - should be
> > easy to make the DB believe it's consistent - export data and rebuild
> > DB. But - no way!!!!
> >
> > And nobody can tell me that there isn't a tool that can analyse the last
> > logfile!!!!
> >
> > But Germany has to contact India. They are NOT allowed to contact
> > Toronto!!
> >
> > Great things happend - received a tool that was supposed to reset the
> > rollforward pending state -> SYS3175
> > Received another copy -> turned out to be AIX executable - does not work
> > on OS/2.
> > Received a new copy - same as first!
> > Received a message to either install C++ (haha) or to use a command line
> > switch
> > Well - guess what did not work -> the command line switch - I received a
> > "syntax error" (although the SYS3175 disappeared when using the option
> > <grin>)
> >
> > The Online Backup as of 15th 10pm did produce the same error.
> >
> > After restoring online backup as of 14th I could do a "db2 rollforward
> > database awt to end of logs and stop".
> >
> > End of logs means: 2000-02-15-15.46.22.000000
> >
> > So I finaly did spend two days on the phone. The application wasn't
> > available during this time - and I lost two days of data.
> >
> > This isn't quiet what a DB is supposed to do during the week.
> >
> > I'm disappointed that these errors happend - but I'm very disappointed
> > that IBM couldn't make it work again! There sure IS someone, who could
> > have done - but they simply were not able to do it!
> >
> > Wolf
Serge Rielau
I don't have any experience in these administrative tasks, so I can't comment
on the particular problem being a DB2 compiler person only - and I shouldn't
anyways.
What I can ensure you is that we kicked of a discussion within service that
tries to evaluate how things could have been done better in your case.
Certainly our support isn't perfect and we have to keep on learning and
improving.
I'm sorry to hear that this event altered your perception of DB2 for future
projects.
It certainly is understandable, though.
Nothing is easier to loose and harder to gain than trust.
Five weeks until EDBT conference in Konstanz :-)
Gruss an den Neckar
Serge
Wolf Duttlinger-Manger
Just to keep you all up to date -
I did receive the patch on saturday - what I was asking for since
Thursday - managed to get the data out of the "bad" online backup and
now am "only" missing the four hours from thursday....