From owner-freebsd-cluster@FreeBSD.ORG Tue Jan 10 17:48:32 2012 Return-Path: Delivered-To: freebsd-cluster@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 4C9F5106564A for ; Tue, 10 Jan 2012 17:48:32 +0000 (UTC) (envelope-from escholtz@argonsoft.de) Received: from coyote.quickmin.net (coyote.quickmin.net [217.14.112.24]) by mx1.freebsd.org (Postfix) with ESMTP id A774D8FC13 for ; Tue, 10 Jan 2012 17:48:31 +0000 (UTC) Received: (qmail 17063 invoked from network); 10 Jan 2012 18:21:49 +0100 Received: by simscan 1.4.0 ppid: 17060, pid: 17061, t: 0.0149s scanners: clamav: 0.97/m:54/d:14286 Comment: DomainKeys? See http://antispam.yahoo.com/domainkeys DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=default; d=argonsoft.de; b=j8lhpOknGy3w7X80sAiBbAkB1ehVVzLv+oBMOWxKUTGObkgpteMPmc24hbloftvEdRczbGuA930wqHJYWxgg/yQ+AkNElFL9AdkPNWZkfWmDOgl95CIjrbM7qxSoeHZtidUC2xzc9ul4SqvEqjqRSnTjoi/uTkNU/wXRwPXFEhE= ; Received: from speedy.quickmin.net (HELO dhcp-172-28-52-194.bauer-de.bauermedia.group) (00000150@217.14.112.11) by coyote.quickmin.net with SMTP; 10 Jan 2012 18:21:49 +0100 Message-ID: <4F0C73AF.5050707@argonsoft.de> Date: Tue, 10 Jan 2012 18:21:51 +0100 From: Erik Scholtz User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:8.0) Gecko/20111105 Thunderbird/8.0 MIME-Version: 1.0 To: freebsd-cluster@freebsd.org Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Subject: HAST - documentation unclear to me X-BeenThere: freebsd-cluster@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Clustering FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 10 Jan 2012 17:48:32 -0000 Hi, again I'm in the need of building a FreeBSD cluster (for loadbalancing purposes, not for failover purposes). Still GFS, GPFS or OCFS2 isn't ported to FreeBSD yet, so I was reading about HAST. The one point I do not get is the following part: "[..] HAST works in Primary-Secondary (Master-Backup, Master-Slave) configuration, which means that only one of the cluster nodes can be active at any given time. Active node will be called Primary node. This is the node that will be able to handle I/O requests to HAST-managed devices. [...]" IMHO this can mean two things: 1) Only the Master can be operational, the slave is in standby (and therefor can not be used as loadbalancing system) 2) Slave can be operational too, but on a write-access to the slave it will be sent to the master, that syncs it back to the slave again (which will cost performance but will work). Does anyone of you have an idea which option (1 or 2) is the way HAST will work? If option 1 is the behaviour of HAST: Any recommendations howto loadbalance an apache webserver (with read/write operations)? Mounting the data from a NFS-Volume costs a lot of time (nearly 1 second slower per request) and therefor is not recommended in an apache highperformance-setup. Thanks for your answers. Greetings, Erik -- My blog: http://blog.elitecoderz.net From owner-freebsd-cluster@FreeBSD.ORG Wed Jan 11 09:14:18 2012 Return-Path: Delivered-To: cluster@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id C4CDD106564A for ; Wed, 11 Jan 2012 09:14:18 +0000 (UTC) (envelope-from escholtz@argonsoft.de) Received: from coyote.quickmin.net (coyote.quickmin.net [217.14.112.24]) by mx1.freebsd.org (Postfix) with ESMTP id 2B0B88FC0A for ; Wed, 11 Jan 2012 09:14:17 +0000 (UTC) Received: (qmail 82558 invoked from network); 11 Jan 2012 09:47:35 +0100 Received: by simscan 1.4.0 ppid: 82555, pid: 82556, t: 0.0155s scanners: clamav: 0.97/m:54/d:14293 Comment: DomainKeys? See http://antispam.yahoo.com/domainkeys DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=default; d=argonsoft.de; b=h4cBLEjmxU4g7V5SG4DKSbi7Paf7rY9zXXC7s+byYllJFoAuRxpJth0ag1l1WGsl9X3RrnvSueSMt5rrmwtvkC7bJW6ABK7b5gg2NnZisqN03QxZ0NWOeGcpSvfawqW7kKRLv37tE8C4t+0F0b26iyJ3hewoNamTwV4uDJ9CeDI= ; Received: from speedy.quickmin.net (HELO dhcp-172-28-52-194.bauer-de.bauermedia.group) (00000150@217.14.112.11) by coyote.quickmin.net with SMTP; 11 Jan 2012 09:47:35 +0100 Message-ID: <4F0D4CA2.9060104@argonsoft.de> Date: Wed, 11 Jan 2012 09:47:30 +0100 From: Erik Scholtz User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:8.0) Gecko/20111105 Thunderbird/8.0 MIME-Version: 1.0 To: cluster@freebsd.org Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Cc: Subject: HAST - documentation unclear to me X-BeenThere: freebsd-cluster@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Clustering FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 11 Jan 2012 09:14:18 -0000 Hi, again I'm in the need of building a FreeBSD cluster (for loadbalancing purposes, not for failover purposes). Still GFS, GPFS or OCFS2 isn't ported to FreeBSD yet, so I was reading about HAST. The one point I do not get is the following part: "[..] HAST works in Primary-Secondary (Master-Backup, Master-Slave) configuration, which means that only one of the cluster nodes can be active at any given time. Active node will be called Primary node. This is the node that will be able to handle I/O requests to HAST-managed devices. [...]" IMHO this can mean two things: 1) Only the Master can be operational, the slave is in standby (and therefor can not be used as loadbalancing system) 2) Slave can be operational too, but on a write-access to the slave it will be sent to the master, that syncs it back to the slave again (which will cost performance but will work). Does anyone of you have an idea which option (1 or 2) is the way HAST will work? If option 1 is the behaviour of HAST: Any recommendations howto loadbalance an apache webserver (with read/write operations)? Mounting the data from a NFS-Volume costs a lot of time (nearly 1 second slower per request) and therefor is not recommended in an apache highperformance-setup. Thanks for your answers. Greetings, Erik -- My blog: http://blog.elitecoderz.net From owner-freebsd-cluster@FreeBSD.ORG Wed Jan 11 09:21:05 2012 Return-Path: Delivered-To: freebsd-cluster@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 73E29106564A for ; Wed, 11 Jan 2012 09:21:05 +0000 (UTC) (envelope-from lukasz@wasikowski.net) Received: from bijou.wasikowski.net (bijou.wasikowski.net [IPv6:2001:808:10f::1]) by mx1.freebsd.org (Postfix) with ESMTP id 2804F8FC13 for ; Wed, 11 Jan 2012 09:21:05 +0000 (UTC) Received: from bijou.wasikowski.net (localhost [127.0.0.1]) by bijou.wasikowski.net (Postfix) with ESMTP id 5CFCD5C081; Wed, 11 Jan 2012 10:21:03 +0100 (CET) X-Virus-Scanned: amavisd-new at wasikowski.net Received: from bijou.wasikowski.net ([127.0.0.1]) by bijou.wasikowski.net (bijou.wasikowski.net [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 8qEg2hV3xc6o; Wed, 11 Jan 2012 10:21:01 +0100 (CET) Received: from [192.168.138.150] (cadera.waw.pl [62.121.127.119]) (using TLSv1 with cipher DHE-RSA-CAMELLIA256-SHA (256/256 bits)) (No client certificate requested) by bijou.wasikowski.net (Postfix) with ESMTPSA id 48D3C5C06F; Wed, 11 Jan 2012 10:21:01 +0100 (CET) Message-ID: <4F0D547C.2010004@wasikowski.net> Date: Wed, 11 Jan 2012 10:21:00 +0100 From: =?UTF-8?B?xYF1a2FzeiBXxIVzaWtvd3NraQ==?= User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:9.0) Gecko/20111222 Thunderbird/9.0.1 MIME-Version: 1.0 To: Erik Scholtz References: <4F0C73AF.5050707@argonsoft.de> In-Reply-To: <4F0C73AF.5050707@argonsoft.de> X-Enigmail-Version: 1.3.4 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: freebsd-cluster@freebsd.org Subject: Re: HAST - documentation unclear to me X-BeenThere: freebsd-cluster@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Clustering FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 11 Jan 2012 09:21:05 -0000 W dniu 2012-01-10 18:21, Erik Scholtz pisze: > 1) Only the Master can be operational, the slave is in standby (and > therefor can not be used as loadbalancing system) > > 2) Slave can be operational too, but on a write-access to the slave it > will be sent to the master, that syncs it back to the slave again (which > will cost performance but will work). > > > Does anyone of you have an idea which option (1 or 2) is the way HAST > will work? > > If option 1 is the behaviour of HAST: > Any recommendations howto loadbalance an apache webserver (with > read/write operations)? Mounting the data from a NFS-Volume costs a lot > of time (nearly 1 second slower per request) and therefor is not > recommended in an apache highperformance-setup. It's number 1. You may try to put application files on local FS (fast access) and user's uploads / files generated by application on NFS or in a database. 1 sec/req is a lot for NFS, probably you are using a slow/saturated network or didn't tweak your NFS configuration. Did you turned off mmap and sendfile in apache when using NFS? If you need performance consider changing apache to nginx, it's much faster. Also, look out these tools, you may find them useful: databases/tarantool databases/memcached net/csync2 -- best regards, Lukasz Wasikowski