Re: [HACKERS] memory problems in copying large table to STDOUT

Martin Weinberg <weinberg@osprey.astro.umass.edu> writes:
> I am attaching the backtrace.  This one simultaneously generated
> this kernel message from the md driver:

> raid0_map bug: hash->zone0==NULL for block 1132810879
> Bad md_map in ll_rw_block

> Definitely a problem but no longer sure if it's the same one . . .
> sigh.  

Looks like it is not the same.  As you can see, the error message that
elog is about to report is "cannot read <block#> of <file>", which isn't
too surprising given the kernel notice:

> #1  0x81160e9 in elog (lev=-1, fmt=0x8146a4b "cannot read block %d of %s")
>     at elog.c:81

If this read failure is reproducible then you will need to get that
taken care of before we can make any progress on the original problem.
But it might be a transient failure --- why don't you just start the
copy over again to see?
        regards, tom lane