[Pgpool-general] vacuum fails due to toast table difference

[louis fridkis]

I am using postgres 8.3.5 & pgpool 2.2.2 with:
    replication_mode = true
    load_balance_mode = true
    replicate_select = false
    reset_query_list = 'ABORT; RESET ALL; SET SESSION AUTHORIZATION DEFAULT'
    master_slave_mode = false
    parallel_mode = false

I run psql -p9999 -ddmaip -c"vacuum full verbose analyze ;"

and get the following failure:

INFO:  vacuuming "pg_toast.pg_toast_58429"
ERROR:  kind mismatch among backends. Possible last query was: "vacuum full
verbose analyze ;" kind details are: 0[N] 1[C]
HINT:  check data consistency among db nodes
server closed the connection unexpectedly
        This probably means the server terminated abnormally
        before or while processing the request.
connection to server was lost

This is because this toast table is only present in the primary:

0: psql -p9999 -ddmaip

dmaip=# \d pg_toast.pg_toast_58429
TOAST table "pg_toast.pg_toast_58429"
   Column   |  Type
------------+---------
 chunk_id   | oid
 chunk_seq  | integer
 chunk_data | bytea

0: psql -p2345 -ddmaip

dmaip=# \d pg_toast.pg_toast_58429
TOAST table "pg_toast.pg_toast_58429"
   Column   |  Type
------------+---------
 chunk_id   | oid
 chunk_seq  | integer
 chunk_data | bytea

1: psql -p 2345

dmaip=# \d pg_toast.pg_toast_58429
Did not find any relation named "pg_toast.pg_toast_58429".

Questions:


   1. Did this happen because the toast tables are not replicated?
   2. Is there a way around this without shutting down this production
   system?
   3. Is there a way to avoid this in future?
   4. Does anyone know what the "N" means in: 0[N] 1[C]?


Thanks,

Lou ...
P Please consider the environment before printing this e-mail or any
attachments.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pgfoundry.org/pipermail/pgpool-general/attachments/20110228/e7a4b573/attachment.html>