On Thu, 9 Feb 2006, Steve Snyder wrote:
Red Hat is enabling the "--enable-truncate" option in their Squid packages. Is the use of this option advisable in my circumstances?
I would advice against using this option. It increases the inode usage considerably, only makes a marginal difference in performance, and opens a race window where cache objects can be partially corrupted causing swapin MD5 failures (should be harmless, but somewhat annoying)
Can someone with experience in these circumstances advise me on the performance impact of using "--enable-truncate" (versus outright file deletion) on this cache configuration?
reiserfs should benefit much less from truncate than ext3, thanks to it's different directory/inode layout. And ext3 should only margianlly benefit from it.
Regards Henrik
