Date: Wed, 08 Jun 2011 15:23:06 +0300 From: Mikolaj Golub <trociny@freebsd.org> To: Victor Balada Diaz <victor@bsdes.net> Cc: stable@freebsd.org Subject: Re: Unusable hastd in FreeBSD 8.2 Message-ID: <86vcwgikb9.fsf@kopusha.home.net> In-Reply-To: <20110606144655.GG6624@equilibrium.bsdes.net> (Victor Balada Diaz's message of "Mon, 6 Jun 2011 16:46:55 %2B0200") References: <20110606144655.GG6624@equilibrium.bsdes.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, On Mon, 6 Jun 2011 16:46:55 +0200 Victor Balada Diaz wrote: VBD> Hello, VBD> Hastd in it's current form is not usable on FreeBSD 8.2-RELEASE or in 8-STABLE. You VBD> can see why in this thread: VBD> http://lists.freebsd.org/pipermail/freebsd-fs/2011-February/010752.html VBD> You can see the committed fix in: VBD> http://svnweb.freebsd.org/base?view=revision&revision=219721 VBD> But it's never been MFCd. Is it possible to MFC it to 8-STABLE and maybe VBD> do an errata notice for RELENG_8_2? Actually, it was MFCed. In r220151. Also, I don't think this is an issue that makes hastd unusable in FreeBSD 8.2 :-). The issue is the following. Before switching the node to primary the failover (third-party) script is checking if secondary process is still alive (assuming that in this case the primary on another node is still alive too) and fails if it is -- some protection against split brain. But before r219721 secondary might not die automatically when primary host was down. This can be workarounded. E.g. by removing the check in the script :-). Or setting net.inet.tcp.keepidle to some small value (e.g. 10 seconds) -- this should make secondary notice that another end is dead after this interval. -- Mikolaj Golub
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?86vcwgikb9.fsf>