Re: Problem loading pg_dump file

"Mason Hale" <masonhale(at)gmail(dot)com> writes:
> Here's the first error I run across:
>
> ***(Single step mode: verify
> command)*******************************************
> COPY blocked_info (id, created_at, reason_code, note, do_count_links) FROM
> stdin;
> ***(press return to proceed or enter x and return to
> cancel)********************
> ERROR: invalid input syntax for integer: "2006-10-09 22:55:58"
> CONTEXT: COPY blocked_info, line 1, column id: "2006-10-09 22:55:58"
>
> The part of the script that is responsible for this error is:
>
> COPY blocked_info (id, created_at, reason_code, note, do_count_links) FROM
> stdin;
> 1 2006-10-09 22:55:58 0 \N \N
> 2 2006-10-09 22:55:58 0 \N \N
> 3 2006-10-09 22:55:58 0 \N \N
> 4 2006-10-09 22:55:58 0 \N \N

That's completely bizarre ... it seems like it's just lost the first
field of the COPY data, which is not a failure I've ever heard of
before. I have no theory about that at the moment.

> In addition to the above, I did a new pg_dump using the -Fc option to
> generate an archive in binary/compressed format.
> ...
> The error message is:
> pg_restore: restoring data for table "rawfeed"
> pg_restore: [custom archiver] could not read data block -- expected 4096,
> got 3448
> pg_restore: *** aborted because of error
>
> Note: this generated file to restore this one table is huge: 9.2G by itself.
> Is there an upper limit to the amount of data copy can load at one time?

There is not supposed to be. However, on some platforms programs need
special build flags to deal with files exceeding 4GB. I'm wondering
if your build of pg_restore is missing largefile support for some
reason. What is the platform here, exactly, and how did you build
or come by your PG installation?

regards, tom lane