"create implicit sequence" crashes postgres - Mailing list pgsql-admin

Hey all,

I've been banging my head against a brick wall on this for ages, and getting
nowhere.
We're trying to move off a single VM running postgres 8.1 to a newer more
powerful pair of VMs with a DRBD mirror for storage and postgres 8.2.
But, every attempt to dump the data from 8.1.x into 8.2.6 fails, and fails
exactly the same way.

This SQL:

CREATE TABLE billing_transaction_items (
    id serial NOT NULL,
    billing_transaction_id integer NOT NULL,
    pricing_term_id integer,
    product_id integer NOT NULL,
    is_free boolean NOT NULL,
    is_visible boolean NOT NULL,
    created_at timestamp without time zone NOT NULL,
    updated_at timestamp without time zone NOT NULL,
    cost numeric(15,4) DEFAULT 0.0 NOT NULL,
    tax numeric(15,4) DEFAULT 0.0 NOT NULL,
    quoted_cost numeric(15,4) DEFAULT 0.0 NOT NULL,
    quoted_tax numeric(15,4) DEFAULT 0.0 NOT NULL
);

And any SQL to create a table with more than 24, or 25 (can't remember off top
of my head) characters long, with a serial column, causes this crash:

Feb  4 17:00:03 mcsql1 postgres[6639]: [23-1] NOTICE:  CREATE TABLE will
create implicit sequence "billing_transaction_items_id_seq" for serial column
Feb  4 17:00:03 mcsql1 postgres[6639]: [23-2]  "billing_transaction_items.id"
Feb  4 17:00:03 mcsql1 postgres[6377]: [4-1] LOG:  server process (PID 6639)
was terminated by signal 11
Feb  4 17:00:03 mcsql1 postgres[6377]: [5-1] LOG:  terminating any other
active server processes
Feb  4 17:00:03 mcsql1 kernel: postmaster[6639] general protection
rip:5555557e7fed rsp:7ffffb246110 error:0
Feb  4 17:00:03 mcsql1 postgres[6377]: [6-1] LOG:  all server processes
terminated; reinitializing
Feb  4 17:00:03 mcsql1 postgres[6640]: [7-1] LOG:  database system was
interrupted at 2008-02-04 16:59:54 GMT
Feb  4 17:00:03 mcsql1 postgres[6640]: [8-1] LOG:  checkpoint record is at
0/4CC150
Feb  4 17:00:03 mcsql1 postgres[6640]: [9-1] LOG:  redo record is at 0/4CC150;
undo record is at 0/0; shutdown TRUE
Feb  4 17:00:03 mcsql1 postgres[6640]: [10-1] LOG:  next transaction ID:
0/630; next OID: 24576
Feb  4 17:00:03 mcsql1 postgres[6640]: [11-1] LOG:  next MultiXactId: 1; next
MultiXactOffset: 0
Feb  4 17:00:03 mcsql1 postgres[6640]: [12-1] LOG:  database system was not
properly shut down; automatic recovery in progress
Feb  4 17:00:03 mcsql1 postgres[6640]: [13-1] LOG:  redo starts at 0/4CC1A0
Feb  4 17:00:03 mcsql1 postgres[6640]: [14-1] LOG:  record with zero length at
0/4F22C0
Feb  4 17:00:03 mcsql1 postgres[6640]: [15-1] LOG:  redo done at 0/4F2280
Feb  4 17:00:03 mcsql1 postgres[6640]: [16-1] LOG:  database system is ready

mcsql1 is fully 64bit.
My 32bit desktop can create the table fine, another 64bit server can also
create the table fine. Create a completely fresh postgres database *not* on
DRBD on mcsql1, and it crashes! reiserfs, ext3, no difference.

The only comonality between those that crash is that they're all 64bit Xen
VMs, so far non-VM servers don't crash. But nothing anywhere I can find even
vaguely hints at problems running Postgres under Xen.
However, the 8.1 server was also a Xen VM, although a 32bit install booted
with a 64bit kernel, enabled someone to create these tables with long names.

Am I being throughly dense and missing the obvious? Am I'm doing something
stupid and un-supported? Or have I found a bug, perhaps?

Thanks

--
Mike Williams