"David Leppik" <dleppik(at)vocalabs(dot)com> writes:
> We are intermittently getting results from now() which are around 10 minutes
> in the future. Most calls return a reasonable value. Because the erroneous
> timestamps are in the future, they cannot be explained by transaction
> delays.
Postgres is just reporting what it got from gettimeofday(), so your beef
is with the kernel (or perhaps with glibc) and/or the hardware you're
using. I think I've heard of kernel bugs causing this type of issue.
regards, tom lane