Re: Re: [COMMITTERS] pgsql: Map basebackup tablespaces using a tablespace_map file

On Fri, Jun 5, 2015 at 9:57 AM, Andrew Dunstan <andrew@dunslane.net> wrote:

On 06/04/2015 11:35 PM, Amit Kapila wrote:

Theoretically, I don't see much problem by changing the checks
way you have done in patch, but it becomes different than what
we have in destroy_tablespace_directories() and it is slightly
changing the way check was originally done in
create_tablespace_directories(), basically original check will try
unlink if lstat returns non-zero return code. If you want to proceed
with the changed checks as in v3, then may be we can modify
comments on top of function remove_tablespace_symlink() which
indicates that it works like destroy_tablespace_directories().

The difference is that here we're getting the list from a base backup and it seems to me the risk of having a file we don't really want to unlink is significantly greater.

Okay, I think I can understand why you want to be cautious for

having a different check for this path, but in that case there is a

chance that recovery might fail when it will try to create a symlink

for that file.  Shouldn't we then try to call this new function only

when we are trying to restore from tablespace_map file and also

is there a need of ifdef S_ISLINK in remove_tablespace_link?