Comments on: Recap of Southeast Linux Fest 2009 http://www.xaprb.com/blog/2009/06/21/recap-of-southeast-linux-fest-2009/ Stay curious! Fri, 10 May 2013 18:25:19 +0000 hourly 1 http://wordpress.org/?v=3.5.1 By: Xaprb http://www.xaprb.com/blog/2009/06/21/recap-of-southeast-linux-fest-2009/#comment-16584 Xaprb Mon, 22 Jun 2009 01:42:11 +0000 http://www.xaprb.com/blog/?p=483#comment-16584 Hi Toru! Thanks for commenting. So in terms of transactional behavior, if you want durability you must commit and then explicitly sync, right?

]]> By: Toru Maesaka http://www.xaprb.com/blog/2009/06/21/recap-of-southeast-linux-fest-2009/#comment-16583 Toru Maesaka Mon, 22 Jun 2009 01:29:39 +0000 http://www.xaprb.com/blog/?p=483#comment-16583 Hi!

FYI, TC does use fsync(3) internally (take a look at tchdbmemsync() in tchdb.c). As you would see, TC’s synchronization mechanism comprises of msync(2) and fsync(2).

In terms of synchronization timing, TC pushes this responsibility to the OS _unless_ the sync call for it’s database (tchdbsync(), tcbdbsync() and etc) is explicitly called by the application.

Well, actually this isn’t 100% correct since TC’s btree database (TCBDB) will call it’s sync mechanism internally if the cache usage exceeds the internal cache capacity.

]]>