On Apr 11, 2013, at 10:45 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Yunong Xiao" <yunong@joyent.com> writes:
>> On Mar 27, 2013, at 12:48 PM, Heikki Linnakangas <hlinnakangas@vmware.co=
m> wrote:
>>> This bug was introduced by commit 8805ff6580621d0daee350826de5211d6bb36=
ec3, in 9.2.2 (and 9.1.7 and 9.0.11), which fixed multiple WAL replay issue=
s with Hot Standby. Before that commit, replaying a heap update didn't try =
to keep both buffers locked at the same time, which is necessary for the co=
rrectness of hot standby. The patch fixed that, but missed releasing the ol=
d buffer in this corner case. I was not able to come up with a scenario wit=
h full_page_writes=3Don where this would fail, but I'm also not 100% sure i=
t can't happen.
>>>=20
>>> I scanned through the commit, and couldn't see any other instances of t=
his kind of a bug. heap_xlog_update is more complicated than other redo fun=
ctions, with all the return statements inside it. It could use some refacto=
ring, but for now, I'll commit the attached small fix.
>>>=20
>>> - Heikki
>>> <fix-heap-update-redo-buffer-leak.patch>
>=20
>> Heikki:
>=20
>> Is it safe to apply the patch you attached? We'd really like to roll out=
this fix to production instead of waiting for this to be released as a pos=
tgres minor version update.
>=20
> Eh? That patch *is* in the latest minor versions.
>=20
> regards, tom lane
Indeed it is. My apologies. Thanks so much for the fix!=
