Den 12/08/14 22.42, Alvaro Herrera skrev:
> behrmann@ndgf.org wrote:
>
>> We have observed two rows with the same primary key in one of our production
>> databases. We don't know how to reproduce this problem, but given that this
>> looks like a corruption bug, I figure you will want to know about this.
>
>
> Can you please install extension pageinspect, then grab the TIDs of both
> copies (select ctid from srmlinkgroup where id = ..), then post the
> output of:
>
> select *
> from heap_page_items(get_raw_page('srmlinkgroup', BLKNO))
> where lp = (OFFNO)
>
> where the BLKNO value is the first half of the TID, and OFFNO is the
> second half of the TID, for each of those tuples?
>
> Please attach the results in text files rather than pasting in the
> email, as whitespace is mangled by your mail program.
>
Please find the result of the above query attached.
I can add that filtering by the id or name fields only gives me one of
the rows (the one corresponding to the first row in the attached file).
I guess that means only that row is linked from the indexes.
Thanks for your support,
/gerd
--
Acting NT1 Area Coordinator, NeIC