Comments on: Attempting to unwind the tangled web of pid file creation. http://ebergen.net/wordpress/2009/12/06/attempting-to-unwind-the-tangled-web-of-pid-file-creation/ You will probably want some waders, a pickaxe, and one of those hats with a light on it before you go in here. Fri, 13 Jan 2012 16:39:29 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Log Buffer #172: a Carnival of the Vanities for DBAs | The Pythian Blog http://ebergen.net/wordpress/2009/12/06/attempting-to-unwind-the-tangled-web-of-pid-file-creation/comment-page-1/#comment-261689 Log Buffer #172: a Carnival of the Vanities for DBAs | The Pythian Blog Fri, 11 Dec 2009 18:56:33 +0000 http://ebergen.net/wordpress/?p=346#comment-261689 [...] long as we’re in /var, here’s Eric Bergen attempting to unwind the tangled web of pid file creation, which he started doing having discovered a pid file creation race condition in [...] [...] long as we’re in /var, here’s Eric Bergen attempting to unwind the tangled web of pid file creation, which he started doing having discovered a pid file creation race condition in [...]

]]> By: Antony Curtis http://ebergen.net/wordpress/2009/12/06/attempting-to-unwind-the-tangled-web-of-pid-file-creation/comment-page-1/#comment-261304 Antony Curtis Mon, 07 Dec 2009 06:19:01 +0000 http://ebergen.net/wordpress/?p=346#comment-261304 Just FYI, __EMX__ denotes OS/2 using the EMX runtime, originally developed by Eberhard Mattes to port unix applications to the OS/2 platform. Just FYI, __EMX__ denotes OS/2 using the EMX runtime, originally developed by Eberhard Mattes to port unix applications to the OS/2 platform.

]]> By: Eric Bergen http://ebergen.net/wordpress/2009/12/06/attempting-to-unwind-the-tangled-web-of-pid-file-creation/comment-page-1/#comment-261292 Eric Bergen Mon, 07 Dec 2009 03:47:42 +0000 http://ebergen.net/wordpress/?p=346#comment-261292 It doesn't fix it. It just makes it far less likely. Innodb initialization can take minutes or even hours to finish depending on how much work it has to do. WIth my patch the pid file is created after args are parsed. In the specific case I ran into while testing some changes to mysqld_safe the race was between me switching terminals to kill mysqld and the pid file being created. I often "won" that race due to the time it takes to initialize innodb. With this patch I can't win because the time before creating the pid file has been significantly reduced. There is also another side case which is allowing two mysqlds to start on the same directly and perform storage engine initialization. Anyone that has had this happen with innodb has seen the flurry of file locking errors innodb spits out when this happens. This patch doesn't fix that race condition either but makes it significantly less likely to happen. It doesn’t fix it. It just makes it far less likely. Innodb initialization can take minutes or even hours to finish depending on how much work it has to do. WIth my patch the pid file is created after args are parsed.

In the specific case I ran into while testing some changes to mysqld_safe the race was between me switching terminals to kill mysqld and the pid file being created. I often “won” that race due to the time it takes to initialize innodb. With this patch I can’t win because the time before creating the pid file has been significantly reduced.

There is also another side case which is allowing two mysqlds to start on the same directly and perform storage engine initialization. Anyone that has had this happen with innodb has seen the flurry of file locking errors innodb spits out when this happens. This patch doesn’t fix that race condition either but makes it significantly less likely to happen.

]]> By: Davi Arnaut http://ebergen.net/wordpress/2009/12/06/attempting-to-unwind-the-tangled-web-of-pid-file-creation/comment-page-1/#comment-261290 Davi Arnaut Mon, 07 Dec 2009 03:30:28 +0000 http://ebergen.net/wordpress/?p=346#comment-261290 So, how does this fix the race that you mentioned? It seems that the patch only makes it more unlikely to happen. So, how does this fix the race that you mentioned? It seems that the patch only makes it more unlikely to happen.

]]>