Re: Plugging fd leaks (was Re: Switching timeline over streaming replication)
Tom Lane <tgl@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Amit Kapila <amit.kapila@huawei.com>
Cc: "'Heikki Linnakangas'" <hlinnakangas@vmware.com>, "'PostgreSQL-development'" <pgsql-hackers@postgresql.org>
Date: 2012-11-26T15:04:10Z
Lists: pgsql-hackers
Amit Kapila <amit.kapila@huawei.com> writes: > On Monday, November 26, 2012 7:01 PM Heikki Linnakangas wrote: >> Hmm, if it's just for locking purposes, how about using a lwlock or a >> heavy-weight lock instead? > Its not only for lock, the main idea is that we create temp file and write > modified configuration in that temp file. > In end if it's success, then we rename temp file to .conf file but if it > error out then at abort we need to delete temp file. > So in short, main point is to close/rename the file in case of success (at > end of command) and remove in case of abort. I'd go with the TRY/CATCH solution. It would be worth extending the fd.c infrastructure if there were multiple users of the feature, but there are not, nor do I see likely new candidates on the horizon. regards, tom lane