Hi Tom, I tried to reproduce the problem once more creating a new blank database and some tables with constraints. I could not reproduce the problem. I then went back and tried to reproduce the problem with the original database. Now I can not reproduce even that although it happened multiple times to me before. I would suggest that I keep an eye on it and try to figure out how to reproduce the problem. Thanks for your time!! Alberto ps. I had switched SELinux off for now. Alberto <nspmma ( at ) freenet ( dot ) de> writes:
I would definetely be willing to create a self contained test case but I am not sure what that would be. I can create a dump file from a test database, but that would be after using pg_dump so you could not check that. Do you want me to create a database, shut it down and zip up the database files? Which files would those be?
Well, basically I want to know how to create a database from which pg_dump can't dump indexes ... can you reconstruct the sequence that got you into this condition? Also ... this is really grasping at straws, but do you have SELinux set in enforcement mode in your RHEL4 system, and if so does changing it to permissive mode make a difference? We've seen a bunch of issues with SELinux :-( --- this does not seem like any of them, but you never know.
Is there a way for me to find out if the dump file created with -Fc does contain information on the indexes?
If "pg_restore -l" shows a lot of stuff but not the indexes (either as indexes or as constraints) then I'd say that's sufficient evidence. regards, tom lane
