Klint Gore <kgore4(at)une(dot)edu(dot)au> writes:
> Tom Lane wrote:
>> Maybe it's
>> dying here after having leaked a lot of memory for some other reason
>> --- try watching the pg_dump process size while it runs.
> Peak memory usage was about 540m which brought the total usage for the
> machine to about half the physical memory allocated (3g total).
Well, that might well explain the failure. pg_dump does suck a lot of
schema information into memory at startup, but 540m seems excessive.
Maybe you've found a memory leak in pg_dump (it wouldn't be the first
one). Does this database have a particularly large number of objects?
regards, tom lane