17cc78ef01
written a generic framework of rules that the contrib makefiles can use instead of writing their own each time. You only need to set a few variables and off you go. |
||
---|---|---|
.. | ||
Makefile | ||
pg_resetxlog.c | ||
README.pg_resetxlog |
pg_resetxlog is a program to clear the WAL transaction log (stored in $PGDATA/pg_xlog/), replacing whatever had been in it with just a dummy shutdown-checkpoint record. It also regenerates the pg_control file if necessary. THIS PROGRAM WILL DESTROY VALUABLE LOG DATA!!! Don't run it unless you really need it!!! pg_resetxlog is primarily intended for disaster recovery --- that is, if your pg_control and/or xlog are hosed badly enough that Postgres refuses to start up, this program will get you past that problem and let you get to your data files. But realize that without the xlog, your data files may be corrupt due to partially-applied transactions, incomplete index-file updates, etc. You should dump your data, check it for accuracy, then initdb and reload. A secondary purpose is to cope with xlog format changes without requiring initdb. To use pg_resetxlog for this purpose, just be sure that you have cleanly shut down your old postmaster (if you're not sure, see the contrib module pg_controldata and run it to be sure the DB state is SHUTDOWN). Then run pg_resetxlog, and finally install and start the new version of the database software. To run the program, make sure your postmaster is not running, then (as the Postgres admin user) do pg_resetxlog $PGDATA As a safety measure, the target data directory must be specified on the command line, it cannot be defaulted. If pg_resetxlog complains that it can't reconstruct valid data for pg_control, you can force it to invent plausible data values with pg_resetxlog -f $PGDATA If this turns out to be necessary then you *definitely* should plan on immediate dump, initdb, reload --- any modifications you do to the database after "pg_resetxlog -f" would be likely to corrupt things even worse.