On Tue, Apr 28, 2015 at 10:54 AM, Alvaro Herrera <alvherre@2ndquadrant.com>
wrote:
> Alvaro Herrera wrote:
>
> > > I think this is confusing:
> > >
> > > +/*
> > > + * Read the offset of the first member of the given multixact.
> > > + */
> > >
> > > This is confusing to me because the two subdirectories of pg_multixact
> > > are called "members" and "offsets". Here you are talking about the
> > > offset of the first member. Maybe I'm just slow, but that seems like
> > > conflating terminology. You end up with a function called
> > > read_offset_for_multi() that is actually looking up information about
> > > members. Ick.
> >
> > Yeah, I introduced the confusing terminology while inventing multixacts
> > initially and have regretted it many times. I will think about a better
> > name for this. (Meanwhile, on IM Robert suggested
> > find_start_of_first_multi_member)
>
> Pushed. I chose find_multixact_start() as a name for this function.
>
Starting with commit b69bf30b9bfacafc733a9ba7 and continuing to this
just-described commit, I can no longer upgrade from a 9.2.10 database using
pg_upgrade.
I can reproduce it from a clean 9.2 install which has never even been
started up.
Deleting files from new pg_multixact/offsets ok
Setting oldest multixact ID on new cluster ok
Resetting WAL archives ok
*failure*
Consult the last few lines of "pg_upgrade_server.log" for
the probable cause of the failure.
The last few lines are:
command: "../bisect/bin/pg_ctl" -w -l "pg_upgrade_server.log" -D
"../data2/" -o "-p 50432 -b -c synchronous_commit=off -c fsync=off -c
full_page_writes=off -c listen_addresses='' -c
unix_socket_permissions=0700 -c
unix_socket_directories='/home/jjanes/pgsql/git'" start >>
"pg_upgrade_server.log" 2>&1
waiting for server to start....LOG: database system was shut down at
2015-04-28 11:08:18 PDT
FATAL: could not access status of transaction 1
DETAIL: Could not open file "pg_multixact/offsets/0000": No such file or
directory.
LOG: startup process (PID 3977) exited with exit code 1
LOG: aborting startup due to startup process failure
Cheers,
Jeff
