From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: From: Philipp Reisner To: drbd-announce@lists.linbit.com Date: Fri, 22 Oct 2010 18:53:11 +0200 MIME-Version: 1.0 Content-Type: Text/Plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Message-Id: <201010221853.11688.philipp.reisner@linbit.com> Cc: drbd-user@lists.linbit.com Subject: [DRBD-announce] drbd-8.3.9.tar.gz Reply-To: drbd-user@lists.linbit.com List-Id: Announcements of new releases and critical bugs found List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Dear DRBD-Users, Here is the final 8.3.9 release, enjoy! There are a few interesting changes in this release. The protocol got a new header type, that allows us to send more than 64kByte of data in one go. As a result, we can now support BIOs up to 128KByte a peace. That helps in high performance setups. Further DRBD got the ability to restart IO requests after IO was suspended. This enables us to have the new configuration directive on-no-data-accessible as well as to restart IO in other circumstances. 8.3.9 (api:88/proto:86-95) -------- * Fix for possible deadlock on IO error during resync * Fixed a race condition between adding and removing network configuration. Lead to a BUG_ON() when triggered. * Fixed spurious full syncs that could happen after an empty resync and concurrent connection loss. * Fixed spurious full syncs that happened when connection got lost while one node was in WFSyncUUID state (Bugz 318) * Fixed a race in the meta-data update code path, that could lead to forgotten updates to the meta-data. That in fact could lead to unexpected behavior at the next connect * Fixed potential deadlock on detach * Fixed potential data divergence after multiple failures * Implicitly create unconfigured devices which are referenced in sync-after dependencies. * OCF RA now also works with pacemaker 1.1 * Allow BIO sizes of up to 128kByte. Note: In case drbd-proxy is used, at least version 1.0.16 of drbd-proxy is required. * New configuration keyword on-no-data-accessible. Possible values io-error, and suspend-io. The default is "io-error", which matches the previous behavior. * If the fencing policy is set to resource-and-stonith, the primary node will creates the new current UUID _after_ the fencing handler returned. (Before it did immediately) * Rewrote the resync speed control loop code. New configuration parameters c-plan-ahead, c-fill-target, c-delay-target, c-max-rate, c-min-rate. * Disable activity log updates when all blocks of an unconnected device is are out of sync. That can be activated by using "invalidate-remote" on an unconnected primary. * Improved IPv6 support: link local addresses * Improved resync speed display in /proc/drbd http://oss.linbit.com/drbd/8.3/drbd-8.3.9.tar.gz http://git.drbd.org/drbd-8.3.git/?a=tag;h=drbd-8.3.9 http://www.linbit.com/support/drbd-8.3.9/ Best, Phil -- : Dipl-Ing Philipp Reisner : LINBIT | Your Way to High Availability : Tel: +43-1-8178292-50, Fax: +43-1-8178292-82 : http://www.linbit.com DRBD(R) and LINBIT(R) are registered trademarks of LINBIT, Austria.