From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Merlin Moncure <mmoncure(at)gmail(dot)com> |
Cc: | Simon Riggs <simon(at)2ndquadrant(dot)com>, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, Jeff Davis <pgsql(at)j-davis(dot)com>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Do we need so many hint bits? |
Date: | 2012-11-20 21:58:46 |
Message-ID: | 15965.1353448726@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Merlin Moncure <mmoncure(at)gmail(dot)com> writes:
> Hm, I wonder if you could squeeze two bits out. ISTM here are the
> interesting cases enumerated:
> 0: xmin unknown
> 1: xmin invalid
> 2: xmin valid, xmax unknown
> 3: xmin valid, xmax invalid
> 4: xmin valid, xmax valid
> Did I miss any?
Yes. xmin unknown, xmax unknown is possible and different from all the
above, ie a tuple can be deleted by the creating transaction. (But it
could still be visible to some of that transaction's snapshots, so you
can't equate this state to "xmin invalid".)
There's a fairly big problem with any of these ideas, and it's not
even on-disk compatibility. It is that we assume that hint bits can be
set without exclusive lock on the buffer. If any of the transitions
xmin unknown -> xmin committed, xmin unknown -> xmin aborted,
xmax unknown -> xmax committed, xmax unknown -> xmax aborted
aren't expressed by setting a bit that wasn't set before, we probably
lose that property, and thereby a whole lot of concurrency.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Jeff Janes | 2012-11-20 22:50:46 | Re: StrategyGetBuffer questions |
Previous Message | Andres Freund | 2012-11-20 21:46:17 | Re: [PATCH] binary heap implementation |