Hi Cedric,

I know the xid wraparound problem. :). 

Right know, I've disabled the slony own vacuum via vac_frequency=0 at
the slony.conf. I know, this is a bad workaround, but it fixed the
problem for me. 

I let autovacuum alone, but I don't know what was the real problem or
why the slony own vac blocks all tables. Is there any way to see / 
debug this? In the logfiles I've seen nothing. 

Also this query doesn't help me to finde the problem:

SELECT dbu.usename as locker, l.mode as locktype,
pg_stat_get_backend_pid(S.backendid) as pid, db.datname||'.'||
n.nspname||'.'||r.relname as relation, l.mode,
substring(pg_stat_get_backend_activity(S.backendid ), 0, 30) as query
FROM pg_user dbu, (SELECT pg_stat_get_backend_idset() AS backendid) AS
S, pg_database db, pg_locks l, pg_class r, pg_namespace n WHERE db.oid =
pg_stat_get_backend_dbid(S.backendid) AND dbu.usesysid =
pg_stat_get_backend_userid(S.backendid) AND l.pid =
pg_stat_get_backend_pid(S.backendid) AND l.relation = r.oid AND
l.database = db.oid AND l.database = db.oid AND l.granted ORDER BY
db.datname, n.nspname, r.relname, l.mode;

Regards,

Martin