Comments on: Tip: faster than TRUNCATE http://code.openark.org/blog/mysql/tip-faster-than-truncate Blog by Shlomi Noach Wed, 01 Feb 2012 20:47:51 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: shlomi http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-59885 shlomi Tue, 22 Nov 2011 07:38:47 +0000 http://code.openark.org/blog/?p=1896#comment-59885 @grace, The RENAME itself should not be affected by number of rows, since it merely maps to a file rename. I say "should" though I've renamed very large tables with no overhead. @grace,
The RENAME itself should not be affected by number of rows, since it merely maps to a file rename. I say "should" though I've renamed very large tables with no overhead.

]]> By: grace http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-59883 grace Tue, 22 Nov 2011 07:22:06 +0000 http://code.openark.org/blog/?p=1896#comment-59883 hey shlomi, just want to ask if number of records in the table affects renaming table performance? hey shlomi,

just want to ask if number of records in the table affects renaming table performance?

]]> By: shlomi http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11334 shlomi Wed, 10 Mar 2010 16:09:17 +0000 http://code.openark.org/blog/?p=1896#comment-11334 @Hans, Well, I would do well with 8-10 seconds. However, as I described, I waited for long minutes. Connections got into Locked state. Nothing would happen (on one occasion I waited for 15 minutes before calling it off). @Hans,
Well, I would do well with 8-10 seconds. However, as I described, I waited for long minutes. Connections got into Locked state. Nothing would happen (on one occasion I waited for 15 minutes before calling it off).

]]> By: Hans-Henrik http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11332 Hans-Henrik Wed, 10 Mar 2010 13:46:38 +0000 http://code.openark.org/blog/?p=1896#comment-11332 The big difference, i think is the time the table 'log_table' is unavailable. If you truncate, the table will be locked until the space has been freed (8-10 seconds for _very_ large tables, if you have IO at the same time). If you do the renaming first, the empty table will appear _very_ fast, while the dropping of the old table does not involve locking of any table that is in use, and so - probably doesn't matter much if it takes a few seconds. The big difference, i think is the time the table 'log_table' is unavailable.

If you truncate, the table will be locked until the space has been freed (8-10 seconds for _very_ large tables, if you have IO at the same time). If you do the renaming first, the empty table will appear _very_ fast, while the dropping of the old table does not involve locking of any table that is in use, and so - probably doesn't matter much if it takes a few seconds.

]]> By: shlomi http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11328 shlomi Wed, 10 Mar 2010 11:34:05 +0000 http://code.openark.org/blog/?p=1896#comment-11328 @Vojtech, Thank you! @Vojtech,
Thank you!

]]> By: Vojtech Kurka http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11327 Vojtech Kurka Wed, 10 Mar 2010 11:30:18 +0000 http://code.openark.org/blog/?p=1896#comment-11327 Morgan, large ibd files are not the only cause. Shlomi, It's a bug, just look at: http://bugs.mysql.com/bug.php?id=35077 http://bugs.mysql.com/bug.php?id=51325 http://bugs.mysql.com/bug.php?id=51557 Morgan, large ibd files are not the only cause. Shlomi, It's a bug, just look at: http://bugs.mysql.com/bug.php?id=35077 http://bugs.mysql.com/bug.php?id=51325 http://bugs.mysql.com/bug.php?id=51557

]]> By: Morgan Tocker http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11315 Morgan Tocker Tue, 09 Mar 2010 23:23:15 +0000 http://code.openark.org/blog/?p=1896#comment-11315 shlomi: The table renamed & dropped still has some contention issues (lock_open mutex will be held), but your trick is still useful for emptying a table and being able to use it again. shlomi: The table renamed & dropped still has some contention issues (lock_open mutex will be held), but your trick is still useful for emptying a table and being able to use it again.

]]> By: shlomi http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11312 shlomi Tue, 09 Mar 2010 14:44:01 +0000 http://code.openark.org/blog/?p=1896#comment-11312 @Darius, Not a problem, since the dropped table has no contention: no one is writing to it. @Darius,

Not a problem, since the dropped table has no contention: no one is writing to it.

]]> By: Darius Jahandarie http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11311 Darius Jahandarie Tue, 09 Mar 2010 14:02:39 +0000 http://code.openark.org/blog/?p=1896#comment-11311 Isn't the table being dropped after renamed though? It should still have the same contention on the disk (if that was the original problem), as long as you are deleting the data. Isn't the table being dropped after renamed though? It should still have the same contention on the disk (if that was the original problem), as long as you are deleting the data.

]]> By: shlomi http://code.openark.org/blog/mysql/tip-faster-than-truncate/comment-page-1#comment-11310 shlomi Tue, 09 Mar 2010 13:19:50 +0000 http://code.openark.org/blog/?p=1896#comment-11310 Hi Morgan, So simple... :D, thanks! At least on two machines I have xfs for sure, though... So there's still a small mystery... Hi Morgan,

So simple... :D , thanks!

At least on two machines I have xfs for sure, though... So there's still a small mystery...

]]>