Re: A Haunted Database

Thanks! Turning off the nightly vacuum script did the trick. Now . . . any
idea why vacuum would be so damaging? It certainly appears, at least for me,
that the routine is more trouble than it is worth. Is it a malfunction that
can be overwritten or a bug or something else?

Again many thanks. I can sleep without fear now

Rob



>Do you have any automated program accessing the database overnight?  IE a
>malfunctioning backup or vacuum script?  You might also want to do a diff
>-C1 first_dump second_dump to see what is actually being changed.
>
>At 11:40 AM 4/8/00, Robert Cleveland wrote:
>>Here's a mystery I hope someone can solve for me.
>>
>>We are entering blocks of HTML into a table called bodyparts. We use PHP3
to
>>break up these blocks into several chunks to keep the length below the
>>maximum. When the end user calls up the section, the "bodyparts" are
>>extracted and re-assembled.
>>
>>The output pages work fine . . . for a while. We set up the output pages
>>during the day, check them for accuracy and go to bed thinking we have
done
>>a great job. Then , in the middle of the night, something happens and when
>>we awake, we find the HTML has been scrambled like so many breakfast eggs.
>>Not all sections are scrambled. In fact it is the same sections every
single
>>time. So we re-enter the data, check it, assume we are done, and then the
>>same thing happens the next day.
>>
>>To gather some empirical evidence, I ran pg_dump at 7pm on the offending
>>table. I check the output pages at midnight the same evening, and they all
>>were good. When I got back in front of the computer at 9am, the pages were
>>scrambled again. I ran pg_dump a second time to a separate file. The file
>>sizes were different (insert scary music here). No one had touched the
>>database or the pages.
>>
>>I reloaded the data and everything is back to normal. But I suspect it
will
>>happen again tonight and I am afraid. Does anyone know what inhuman entity
>>might be causing this to occur?
>>
>>