Обсуждение: Corrupted index file after restoring WAL on warm spare server

Поиск
Список
Период
Сортировка

Corrupted index file after restoring WAL on warm spare server

От
"Michael Nolan"
Дата:
Recently I went through the steps to refresh the database (8.3.4) on my development server (running Fedora Core 5 Linux),
making a tarball of the live database, then restoring it on the development server, and running all the archived WAL files.

Everything worked fine as far as I can tell, I don't see any errors in any of my logs.

Howevever, when I go to access one table, the index appears to be corrupted because the record I get for one query doesn't match the record key I give.

The search works correctly on the live server.

Reindexing that table on the warm spare system fixes the problem.

I redid setting up the warm spare server a second time to see if it was a one-time anomaly, but I have the exact same condition again.

What could be causing this?  If this is a bug, what can I do to help track it down?
--
Mike Nolan
htfoot@gmail.com

Re: Corrupted index file after restoring WAL on warm spare server

От
"Michael Nolan"
Дата:
Sorry, I meant 8.2.4 (darn typo virus)

On 5/24/07, Michael Nolan <htfoot@gmail.com> wrote:
Recently I went through the steps to refresh the database (8.3.4) on my development server (running Fedora Core 5 Linux),
making a tarball of the live database, then restoring it on the development server, and running all the archived WAL files.

Everything worked fine as far as I can tell, I don't see any errors in any of my logs.

Howevever, when I go to access one table, the index appears to be corrupted because the record I get for one query doesn't match the record key I give.

The search works correctly on the live server.

Reindexing that table on the warm spare system fixes the problem.

I redid setting up the warm spare server a second time to see if it was a one-time anomaly, but I have the exact same condition again.

What could be causing this?  If this is a bug, what can I do to help track it down?
--
Mike Nolan
htfoot@gmail.com

Re: Corrupted index file after restoring WAL on warm spare server

От
Tom Lane
Дата:
"Michael Nolan" <htfoot@gmail.com> writes:
> Howevever, when I go to access one table, the index appears to be corrupted
> because the record I get for one query doesn't match the record key I give.

> Reindexing that table on the warm spare system fixes the problem.

> I redid setting up the warm spare server a second time to see if it was a
> one-time anomaly, but I have the exact same condition again.

> What could be causing this?  If this is a bug, what can I do to help track
> it down?

It sounds like you have a reproducible test case --- can you make it
available to someone else?

            regards, tom lane

Re: Corrupted index file after restoring WAL on warm spare server

От
"Michael Nolan"
Дата:
On 5/24/07, Tom Lane <tgl@sss.pgh.pa.us> wrote:

It sounds like you have a reproducible test case --- can you make it
available to someone else?

Not very easily, Tom.  The tarball files are around 35 GB (uncompressed) and I assume they'd only work on a fairly similar system anyway.

I assume shell level access would be needed. I may be able to set up a way to get to someone through the firewall, but there are some security issues I'd have to clear with our Executive Director, since there's quite a bit of confidential information in the database and elsewhere on the server. 

Is this something where I could call someone and have them walk me through checking things?

I tried the same thing on a 3rd server (similar hardware and same O/S), and the query is working correctly there.  I'm not sure what that tells me.
--
Mike Nolan