I think I coded a very basic version of the UNLOGGED to LOGGED patch
(only wal_level=minimal case for the moment).

To remove the INIT fork, I changed somehow PendingRelDelete to have
a flag "bool onlyInitFork" so that the delete would remove only the INIT
fork at commit.

Everything "works" (note the quotes...) except in the case of not-clean
shutdown ("-m immediate" to pg_ctl stop). The reason is that the replay
code doesn't have any idea that it has to remove only the INIT fork: it will
remove ALL forks; so at the end of the redo procedure (at startup, after
a "pg_ctl -m immediate stop") the table doesn't have any forks at all.

See xact_redo_commit:

/* Make sure files supposed to be dropped are dropped */
for (i = 0; i < xlrec->nrels; i++)
   for (fork = 0; fork <= MAX_FORKNUM; fork++)
                  if (smgrexists(srel, fork))
                             XLogDropRelation(xlrec->xnodes[i], fork);
                             smgrdounlink(srel, fork, true);

Should I change xl_xact_commit in order to have, instead of:

/* Array of RelFileNode(s) to drop at commit */
RelFileNode xnodes[1];          /* VARIABLE LENGTH ARRAY */

an array of structures such as:

    RelFileNode   relfilenode;
    bool               onlyInitFork;


Otherwise I don't know how to tell the redo commit code to delete only
the INIT fork, instead of all the relation forks...
(maybe I'm doing all wrong: I'm open to any kind of suggestion here...)


Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:

Reply via email to