SegmentMap is corrupted

[Steve Albin]

On Tuesday, February 13, 2007 sunshine at public.kherson.ua (Ruslan Zasukhin) wrote:

>> Is the segmentmap corruption something that can happen due to a bug in my app?
>> What kind of thing could cause this?
>
>This error can be expected in case of many new WRITE into db without Flush,
>Then interrupt application (crash, system failure).

This is interesting.  This database had very few writes done to it except for the Convert_1_2.  My old app used Valentina 1 and this user has been using it for years with no problem.  I am trying to upgrade my app to use Valentina 2, but so far without success for this user.  (You can see my Mantis reports under id steveja7.)   At this point, he is beta testing the app.  Data entry for this app is primarily record by record from user data entry.  The only high volume writing to the DB would be the convert.  Also, he has not reported any hard crashes except for the one resulting from the segmentmap corruption.

Also, this user has, by far, the largest database of any of my users. So there are many more writes done on the conversion than for my other users, all of whom have not reported this problem.  (I only have about a half dozen users for this version - not many I admit.) 

We have had a segmentmap corruption previously that was isolated to adding a new VText field to a database converted from Val 1 to Val 2.  That has been fixed.  Now I'm wondering if there is some other error resulting from the conversion process.  

>> When this error occurs, what is the prescribed method of trying to rescue the
>> database?  Reindex? Clone DB?
>
>clone

Did not work for me, but I am going to try it again.  Maybe there was something wrong with the way I did it as it took 4 hours.
>
>If .ind file is corrupted, then yes REINDEX can help,
>Even better trash .ind file and open db again.

Tried this.  Didn't help.

>
>You can try DIAGNOSE db VERYHIGH.

Did this.  No errors reported.


>
>> I tried cloning this DB using VStudio and it took 4 hours.  This is a 40MB
>> database.  Is that an expected time frame for cloning a DB that size?
>
>Seems very long. If you can, compressed db and put on some FTP for me.
>We will profile later.

This database is available and mentioned with a link to it on Mantis report number 2189.  


>
>> I'm suspect of that clone function since I returned the DB to my user and the
>> segment error still occurred.
>
>Then you need ASK your user what steps he do.
>And we need try reproduce this issue.
>
I have repeated the steps my user has done, but it won't crash for me.   But, I know it is crashing for him because he is sending me the V4RB_Logs that I have turned on for his app.  I think it may be some internal timing issue that I can't reproduce.  He is on a fast Windows machine and I am using Virtual PC.  My Mantis report has the small project that he has used to reproduce the error with his database.  

It is also difficult to work on this problem with my user being hundreds of miles away.  I have to rely on him telling me accurately what he is doing.  Not always an easy task.  But, with this information about segmentmap errors, I will try to step back and recreate everything with this in more focus.

Thanks,

Steve



 
-- 
Steve Albin - Montclair, NJ
http://www.jazzdiscography.com/
http://www.steve-albin.com/