Обсуждение: Issues with restoring

Hello Folx!

We are running into a problem that I wanted to run by the group before I report it as a bug.

We have a medium sized database that when dumped creates +4G files within the tar archive.  When we restore it seems
thatpg_restore has a 4G limit for reading files, once it reads 4G of a file, it moves on to the next file.  Has anyone
elseexperienced this problem?  Is there a documented way for mitigating this issue that I have not found?  Is this a
bugor are we doing something incorrectly to cause this?  It seems to me if pg_restore has a hard limit of 4G filesize
thenpg_dump should have the same limit. 

TIA

-bill

"beer" <beer@cmu.edu> writes:
> We have a medium sized database that when dumped creates +4G files
> within the tar archive.  When we restore it seems that pg_restore has
> a 4G limit for reading files, once it reads 4G of a file, it moves on
> to the next file.  Has anyone else experienced this problem?

There is a member size limit inherent to the tar-archive code, although
I thought it was 8G not 4G.  I'd recommend using custom format (-Fc not
-Ft).

Still, if the thing is truncating your data and not telling you so,
that'd qualify as a bug.

On some platforms there might be a problem with lack of large-file
support at the stdio level, too.  What is your platform?

            regards, tom lane

Tom

Dont think its large file support, I'm running on FC4, the problem existed on FC1 too.

I'll try the custom format and see if that helps.  Thanks for the recommendation.

-bill

On Mon, March 27, 2006 4:35 pm, Tom Lane said:
> "beer" <beer@cmu.edu> writes:
>> We have a medium sized database that when dumped creates +4G files
>> within the tar archive.  When we restore it seems that pg_restore has a
>> 4G limit for reading files, once it reads 4G of a file, it moves on to
>> the next file.  Has anyone else experienced this problem?
>
> There is a member size limit inherent to the tar-archive code, although I
> thought it was 8G not 4G.  I'd recommend using custom format (-Fc not
> -Ft).
>
> Still, if the thing is truncating your data and not telling you so, that'd
> qualify as a bug.
>
> On some platforms there might be a problem with lack of large-file support
> at the stdio level, too.  What is your platform?
>
> regards, tom lane
>
>