I just got done upgrading 2 databases from 7.1.x to 7.2.3. In both cases
the procedure outlined in pg_upgrade.1 failed. In one case, the failure
was catastrophic. In neither case was any data lost (because I backed up
with pg_dumpall first), but in both cases it appears the failure was
similar: Everything appeared to go fine until the first 'vacuum analyze'
after the procedure was complete.
In the catastrophic case, unfortunately, I don't have a lot of info. I
just sort of told myself, "Oh well. Good thing I backed it up," did
another initdb and restored the backup. This was the first case, and the
first time I ever attempted a pg_upgrade.
In the second case, I was a lot more careful. Even so, the first vacuum
analyze failed, but the database seems to have recovered without
incident. Here's what the log said:
DEBUG: database system is ready
DEBUG: --Relation pg_type--
DEBUG: Pages 3: Changed 0, Empty 0; Tup 170: Vac 0, Keep 0, UnUsed 0.
Total CPU 0.00s/0.00u sec elapsed 0.00 sec.
DEBUG: Analyzing pg_type
FATAL 2: open of /home/pgsql/data/pg_clog/0004 failed: No such file or
directory
DEBUG: server process (pid 57274) exited with exit code 2
DEBUG: terminating any other active server processes
DEBUG: all server processes terminated; reinitializing shared memory
and semaphores
DEBUG: database system was interrupted at 2002-10-29 06:19:28 PST
DEBUG: checkpoint record is at 0/4E00093C
DEBUG: redo record is at 0/4E00093C; undo record is at 0/0; shutdown TRUE
DEBUG: next transaction id: 4488258; next oid: 16781
DEBUG: database system was not properly shut down; automatic recovery
in progress
DEBUG: redo starts at 0/4E00097C
DEBUG: clog file /home/pgsql/data/pg_clog/0004 doesn't exist, reading
as zeroes
DEBUG: ReadRecord: record with zero length at 0/4E000B34
DEBUG: redo done at 0/4E000B10
DEBUG: database system is ready
At this point, re-performing 'vacuum analyze' succeeded. Poking around
at the actual data shows no aparent data loss.
1. What happened?
2. Is this an expected side effect of pg_upgrade?
3. Am I correct in believing that no data was actually lost?
4. Did I forget to do something critical before the pg_upgrade -1?