Re: BUG #11264: Auto vacuum wraparound job blocking everything

Hi Alvaro,

We did manual vacuum and reindex of room_contract_service_code and it finis=
hed quickly.  We still have auto vacuum disabled.


1 ) select oid::regclass, relfrozenxid, relminmxid from pg_class where reln=
ame =3D 'room_contract_service_code'; select datname, datfrozenxid, datminm=
xid from pg_database where datname =3D 'jetblue';

 jetblue=3D# select oid::regclass, relfrozenxid, relminmxid from pg_class w=
here relname =3D 'room_contract_service_code'
jetblue-# ;
            oid             | relfrozenxid | relminmxid
----------------------------+--------------+------------
 room_contract_service_code |    953567787 |      20784
(1 row)


jetblue=3D# select datname, datfrozenxid, datminmxid from pg_database where=
 datname =3D 'jetblue';
 datname | datfrozenxid | datminmxid
---------+--------------+------------
 jetblue |    870700471 |          1
(1 row)


2) Latest checkpoint's NextMultiXactId:  23431                             =
                                                                           =
                                                               =20
Latest checkpoint's NextMultiOffset:  5678                                 =
                                                                           =
                                                            =20
Latest checkpoint's oldestMultiXid:   1                                    =
                                                                           =
                                                            =20
Latest checkpoint's oldestMulti's DB: 16423                                =
                                                                           =
                                                            =20

It says the oldest multi is 1, so the database should not have any values b=
etween 1 and 23431 that correspond to pg_upgrade'd multixact values ... so =
what is going on here?  Unless the recent mucking with pg_upgrade to handle=
 multixact's got something wrong.

We used psotgres 9.3.4 to pg_upgrade from 9.1. We just upgraded binaries to=
 9.3.5 just last week ? Would this cause any issues with MultiXid issue.

Thanks
Dinesh