Tom Lane írta:
> Zoltan Boszormenyi <zb@cybertec.at> writes:
>
>> The realtime trace I captured from the hung INSERT shows that it
>> enters two functions repeatedly: _bt_relandgetbuf() and _bt_compare().
>> The pattern in which these functions entered match either _bt_moveright() or
>> _bt_insertonpg().
>>
>
> What that sounds like to me is a corrupt index (infinite loop of
> right-links, perhaps). Have you tried REINDEXing the table?
>
> regards, tom lane
>
No, TRUNCATE was faster because there were no live records in the table.
How can such an infinite loop sneak into an index?
Hardware is from Sun, not a grocery store PC, so I don't suppose it to
be faulty.
Is there anything in the 8.0.x series that fixes this (or a similar) bug?
If I could point to something in the release notes, I may get them to
upgrade
and they may upgrade to a newer generation even.
Thanks.
--
----------------------------------
Zoltán Böszörményi
Cybertec Schönig & Schönig GmbH
http://www.postgresql.at/