Thread: invalid page header problem

invalid page header problem

From
"Tarabas (Manuel Rorarius)"
Date:
Hi!

  We currently have a serious problem with our postgres-DB!

  We are using postgres now for some time and lately we expericened
  that tables were corrupted on the 7.3.2 Database. After recreating
  the Tables by copying the contents to a save-table and then dropping
  and recreating the original table we could at least help that
  withourt losing data.

  We then decided it couldn't hurt to upgrade to 7.4 and hope that the
  problems were solved, but they weren't they just changed. In fact
  now we are losing data because we cannot select * from the table
  anymore without getting the new Error.

  At current we get the Error
  "invalid page header in block 2617 of  relation "xxxx"; zeroing out
  page"

  We actually tried to find out what the problem was and the table was
  fine till some tuple crashed with the bug. We then made a
  pg_filedump, and added "zero_damaged_pages" to the config so that we
  could make a clean dump ... since the table is mainly used for
  high-traffic messaging it was ok to maybe lose the one block of
  messages by zeroing them out.

  The problem is, that this keeps happening and we don't know why ...
  Actually this is a fresh installation of postgres 7.4 and a fresh
  initialized database with a dump inserted with psql -e < dumpfile,
  showing the problems after just 12 hours of work.

  The only thing changed before compiling the postgres is the
  INDEX_MAX_ARGS to make it possible to send more than 32 parameters
  to a function which work and always worked fine because we do inserts
  and updates encapsulated in stored procedures.

  We also tried to check the HDD and could not find any evidence of
  corruption or problems on the HDD-Side ...

  Any help with how to find out more WHY this problem keeps showing up
  and how to fix this is appreciated, because the only thing we see is
  that perhaps some piece of hardware may be corrupt and we have to
  change that, but before spending a lot of money on new hardware i'd
  like to rule out that this is a software-problem!

Regards ... Manuel ...

Re: invalid page header problem

From
Tom Lane
Date:
"Tarabas (Manuel Rorarius)" <tarabas@tarabas.de> writes:
>   We currently have a serious problem with our postgres-DB!

I think you've got hardware problems.  Have you tried running memory and
disk tests?  memtest86 and badblocks are the commonly recommended
tools.  Past reports have suggested that badblocks is more likely to
detect problems if you use the "destructive test" option.

            regards, tom lane

unsubscribe me please!

From
yingluo@umich.edu
Date:
Please take me off the email list. Thank you very much!