From mmonge at gmail.com Tue Jan 25 13:01:42 2005 From: mmonge at gmail.com (Marcos Monge) Date: Tue, 25 Jan 2005 14:01:42 +0100 Subject: Cluster without shared disk, or over shared over NFS Message-ID: <48e9aa4e05012505012909d7a@mail.gmail.com> Hi It's possible to run a redhat cluster without the shared disk in a RHEL 3.0? or do the share using an remote NFS volumen? In my case, I have 2 servers, and a NetApp Filer serving NFS file systems. There is any way to put the shared disk in the nfs? Or not use it at all. I have read in the archives a mail from Lon Hohberger telling that it's posible in RHEL 3.0, but not supported: https://www.redhat.com/archives/piranha-list/2004-April/msg00002.html Any idea about how to do this? Marcos From lhh at redhat.com Tue Jan 25 14:42:54 2005 From: lhh at redhat.com (Lon Hohberger) Date: Tue, 25 Jan 2005 09:42:54 -0500 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <48e9aa4e05012505012909d7a@mail.gmail.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> Message-ID: <1106664174.16910.10.camel@ayanami.boston.redhat.com> On Tue, 2005-01-25 at 14:01 +0100, Marcos Monge wrote: > I have read in the archives a mail from Lon Hohberger telling that > it's posible in RHEL 3.0, but not supported: > https://www.redhat.com/archives/piranha-list/2004-April/msg00002.html As of U3, you can use iSCSI or GNBD without any clever tricks (and I believe it's supported now). At the time the above was written, that wasn't the case. You should also be able to use block devices via iSCAI and/or GNBD for service data as well. The rgmanager code (targetted for RHEL4) doesn't need shared storage to operate internally (nor will it use it...). It can use NFS as data stores for services instead of being limited to SANs/iSCSI/GNBD (obvious limitations apply). (Warning: hyper-long-ugly-url below) http://sources.redhat.com/cgi- bin/cvsweb.cgi/cluster/rgmanager/src/resources/netfs.sh?rev=1.1&content- type=text/x-cvsweb-markup&cvsroot=cluster -- Lon From mmonge at gmail.com Tue Jan 25 18:23:21 2005 From: mmonge at gmail.com (Marcos Monge) Date: Tue, 25 Jan 2005 19:23:21 +0100 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <1106664174.16910.10.camel@ayanami.boston.redhat.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> Message-ID: <48e9aa4e05012510233a44de38@mail.gmail.com> On Tue, 25 Jan 2005 09:42:54 -0500, Lon Hohberger wrote: > As of U3, you can use iSCSI or GNBD without any clever tricks (and I > believe it's supported now). At the time the above was written, that > wasn't the case. You should also be able to use block devices via iSCAI > and/or GNBD for service data as well. I'm thinking in use iSCSI, exported from the NetApp filer. I think this will be the best solution with RHEL3, because I have the same problem with the service data partition (I think that actual cluster suite don't support service data over nfs). The RHEL3 I have installed in this servers is Update 3. But I haved read in someplace that to support iSCSI you need Update 4. Are you sure that U3 support iSCSI? This will be great ;) > The rgmanager code (targetted for RHEL4) doesn't need shared storage to > operate internally (nor will it use it...). It can use NFS as data > stores for services instead of being limited to SANs/iSCSI/GNBD (obvious > limitations apply). hmmm... This means that RHEL4 cluster suite will support all the things I need? ;) (that it's: no shared disk between servers, and aplication service data over nfs) Wouu... ;) I think I will try tomorrow with iSCSI, and maybe reinstall all the things in 1 month when rhel4 is avalaible (14-feb I think). Thanks Marcos > > (Warning: hyper-long-ugly-url below) > > http://sources.redhat.com/cgi- > bin/cvsweb.cgi/cluster/rgmanager/src/resources/netfs.sh?rev=1.1&content- > type=text/x-cvsweb-markup&cvsroot=cluster > > -- Lon > > From lhh at redhat.com Tue Jan 25 20:43:50 2005 From: lhh at redhat.com (Lon Hohberger) Date: Tue, 25 Jan 2005 15:43:50 -0500 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <48e9aa4e05012510233a44de38@mail.gmail.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> <48e9aa4e05012510233a44de38@mail.gmail.com> Message-ID: <1106685830.16910.22.camel@ayanami.boston.redhat.com> On Tue, 2005-01-25 at 19:23 +0100, Marcos Monge wrote: > The RHEL3 I have installed in this servers is Update 3. But I haved > read in someplace that to support iSCSI you need Update 4. Are you > sure that U3 support iSCSI? This will be great ;) You're probably right. GNBD was Red Hat GFS (around U3 timeframe); iSCSI was U4. In any case, make sure you're running the latest Red Hat Cluster Suite 3 errata as well. Bugfixes are good. -- Lon From mmonge at gmail.com Wed Jan 26 13:42:07 2005 From: mmonge at gmail.com (Marcos Monge) Date: Wed, 26 Jan 2005 14:42:07 +0100 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <1106685830.16910.22.camel@ayanami.boston.redhat.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> <48e9aa4e05012510233a44de38@mail.gmail.com> <1106685830.16910.22.camel@ayanami.boston.redhat.com> Message-ID: <48e9aa4e050126054216c33e39@mail.gmail.com> On Tue, 25 Jan 2005 15:43:50 -0500, Lon Hohberger wrote: > > > The RHEL3 I have installed in this servers is Update 3. But I haved > > read in someplace that to support iSCSI you need Update 4. Are you > > sure that U3 support iSCSI? This will be great ;) > > You're probably right. GNBD was Red Hat GFS (around U3 timeframe); > iSCSI was U4. There is any easy way to upgrade from update 3 to update 4? use up2date to install the latest updates/erratas is sufficient to have the iSCSI features? > In any case, make sure you're running the latest Red Hat Cluster Suite 3 > errata as well. Bugfixes are good. Yes, I'm using latest version. Best regards From lhh at redhat.com Wed Jan 26 18:51:25 2005 From: lhh at redhat.com (Lon Hohberger) Date: Wed, 26 Jan 2005 13:51:25 -0500 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <48e9aa4e050126054216c33e39@mail.gmail.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> <48e9aa4e05012510233a44de38@mail.gmail.com> <1106685830.16910.22.camel@ayanami.boston.redhat.com> <48e9aa4e050126054216c33e39@mail.gmail.com> Message-ID: <1106765485.16910.74.camel@ayanami.boston.redhat.com> On Wed, 2005-01-26 at 14:42 +0100, Marcos Monge wrote: > There is any easy way to upgrade from update 3 to update 4? use > up2date to install the latest updates/erratas is sufficient to have > the iSCSI features? Should be. -- Lon From mmonge at gmail.com Thu Jan 27 20:12:08 2005 From: mmonge at gmail.com (Marcos Monge) Date: Thu, 27 Jan 2005 21:12:08 +0100 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <1106765485.16910.74.camel@ayanami.boston.redhat.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> <48e9aa4e05012510233a44de38@mail.gmail.com> <1106685830.16910.22.camel@ayanami.boston.redhat.com> <48e9aa4e050126054216c33e39@mail.gmail.com> <1106765485.16910.74.camel@ayanami.boston.redhat.com> Message-ID: <48e9aa4e05012712123f97c0a0@mail.gmail.com> Hi I just installed and configured the Redhat Cluster Suite, under RHEL 3.0 update 3, with a Network appliance (netapp) filer. This is the step by step: 1 - update the kernel package of RHEL 3.0 to latest avalaible, and reboot with this kernel (if you are using RHEL 3.0 update 4, this step is not necesary). 2 - install the iSCSI-initiator-tools package from rhn. 3 - Configure netapp to export 1 disk by iSCSI to the IQN of the 2 cluster nodes. You can see how to do this in the Netapp Documentation. To obtain the iqn of each linux, start the iscsi driver module, and execute iscsi-iname command: /etc/init.d/iscsi start iscsi-iname 4 - Restart iscsi module to see the new exported disk. You must hava a new device called /dev/sda. You can see the kernel message in console or with dmesg command that tell you the name of the dispositive (if you have other local scsi disk, can be antoher leter, like /dev/sdc, for example). 5 - Make two partitions for the raw shared disks. Use fdisk /dev/sda, or parted /dev/sda, as you prefer. You must create 2 small partitions (10 to 30 mb for each is sufficient). 6 - Make the partitions that you need for the aplication/service data, and format it with mke2fs -j /dev/sdaX (X is the partition number). 7 - Configure the raw partitions, adding in /etc/sysconfig/rawdevice two lines like: /dev/raw/raw1 /dev/sda1 /dev/raw/raw2 /dev/sda2 and restart rawpartitions with /etc/init.d/rawpartitions restart 8 - Continue with the standard installation described in the redhat cluster suite manual guide. I have do all this, and I must say that the cluster is working very well. The switching speed is very fast, and the general performance is also very good. I hope this guide can help somebody now or in the future ;) Best Regards Marcos On Wed, 26 Jan 2005 13:51:25 -0500, Lon Hohberger wrote: > On Wed, 2005-01-26 at 14:42 +0100, Marcos Monge wrote: > > > There is any easy way to upgrade from update 3 to update 4? use > > up2date to install the latest updates/erratas is sufficient to have > > the iSCSI features? > > Should be. > > -- Lon > > From mmonge at gmail.com Thu Jan 27 20:23:39 2005 From: mmonge at gmail.com (Marcos Monge) Date: Thu, 27 Jan 2005 21:23:39 +0100 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <48e9aa4e05012712123f97c0a0@mail.gmail.com> References: <48e9aa4e05012505012909d7a@mail.gmail.com> <1106664174.16910.10.camel@ayanami.boston.redhat.com> <48e9aa4e05012510233a44de38@mail.gmail.com> <1106685830.16910.22.camel@ayanami.boston.redhat.com> <48e9aa4e050126054216c33e39@mail.gmail.com> <1106765485.16910.74.camel@ayanami.boston.redhat.com> <48e9aa4e05012712123f97c0a0@mail.gmail.com> Message-ID: <48e9aa4e05012712235fcc47e2@mail.gmail.com> Sorry, there is an error in the step 3. To obtain the IQN (iscsi uniq identifier) of each linux server, you must start the /etc/init.d/iscsi, and look the /etc/initiatorname.iscsi file, that have the iqn asigned to the linux server. Best regards marcos On Thu, 27 Jan 2005 21:12:08 +0100, Marcos Monge wrote: > Hi > > I just installed and configured the Redhat Cluster Suite, under RHEL > 3.0 update 3, with a Network appliance (netapp) filer. This is the > step by step: > > 1 - update the kernel package of RHEL 3.0 to latest avalaible, and > reboot with this kernel (if you are using RHEL 3.0 update 4, this step > is not necesary). > 2 - install the iSCSI-initiator-tools package from rhn. > 3 - Configure netapp to export 1 disk by iSCSI to the IQN of the 2 > cluster nodes. You can see how to do this in the Netapp Documentation. > To obtain the iqn of each linux, start the iscsi driver module, and > execute iscsi-iname command: > /etc/init.d/iscsi start > iscsi-iname > 4 - Restart iscsi module to see the new exported disk. You must hava a > new device called /dev/sda. You can see the kernel message in console > or with dmesg command that tell you the name of the dispositive (if > you have other local scsi disk, can be antoher leter, like /dev/sdc, > for example). > 5 - Make two partitions for the raw shared disks. Use fdisk /dev/sda, > or parted /dev/sda, as you prefer. You must create 2 small partitions > (10 to 30 mb for each is sufficient). > 6 - Make the partitions that you need for the aplication/service data, > and format it with mke2fs -j /dev/sdaX (X is the partition number). > 7 - Configure the raw partitions, adding in /etc/sysconfig/rawdevice > two lines like: > /dev/raw/raw1 /dev/sda1 > /dev/raw/raw2 /dev/sda2 > and restart rawpartitions with /etc/init.d/rawpartitions restart > 8 - Continue with the standard installation described in the redhat > cluster suite manual guide. > > I have do all this, and I must say that the cluster is working very > well. The switching speed is very fast, and the general performance is > also very good. > > I hope this guide can help somebody now or in the future ;) > > Best Regards > Marcos > > > On Wed, 26 Jan 2005 13:51:25 -0500, Lon Hohberger wrote: > > On Wed, 2005-01-26 at 14:42 +0100, Marcos Monge wrote: > > > > > There is any easy way to upgrade from update 3 to update 4? use > > > up2date to install the latest updates/erratas is sufficient to have > > > the iSCSI features? > > > > Should be. > > > > -- Lon > > > > > From bbeaver at pathfire.com Thu Jan 27 20:28:34 2005 From: bbeaver at pathfire.com (Brian Beaver) Date: Thu, 27 Jan 2005 15:28:34 -0500 Subject: Cluster without shared disk, or over shared over NFS Message-ID: <6204EFD303E1D24297BBD43A6ED722E7219882@RSWMS02> Thanks for the good info. So to summarize, Redhat clustering still appears to require "local" disks. The Netapp iSCSI solution just helps the drives appear local to the O/S. Any idea if Redhat clustering is allowed without the shared drive requirement, similar to Veritas clustering using multiple NIC's for heartbeats instead of a shared drive partition? -----Original Message----- From: piranha-list-bounces at redhat.com [mailto:piranha-list-bounces at redhat.com] On Behalf Of Marcos Monge Sent: Thursday, January 27, 2005 3:24 PM To: Piranha clustering/HA technology Subject: Re: Cluster without shared disk, or over shared over NFS Sorry, there is an error in the step 3. To obtain the IQN (iscsi uniq identifier) of each linux server, you must start the /etc/init.d/iscsi, and look the /etc/initiatorname.iscsi file, that have the iqn asigned to the linux server. Best regards marcos On Thu, 27 Jan 2005 21:12:08 +0100, Marcos Monge wrote: > Hi > > I just installed and configured the Redhat Cluster Suite, under RHEL > 3.0 update 3, with a Network appliance (netapp) filer. This is the > step by step: > > 1 - update the kernel package of RHEL 3.0 to latest avalaible, and > reboot with this kernel (if you are using RHEL 3.0 update 4, this step > is not necesary). > 2 - install the iSCSI-initiator-tools package from rhn. > 3 - Configure netapp to export 1 disk by iSCSI to the IQN of the 2 > cluster nodes. You can see how to do this in the Netapp Documentation. > To obtain the iqn of each linux, start the iscsi driver module, and > execute iscsi-iname command: > /etc/init.d/iscsi start > iscsi-iname > 4 - Restart iscsi module to see the new exported disk. You must hava a > new device called /dev/sda. You can see the kernel message in console > or with dmesg command that tell you the name of the dispositive (if > you have other local scsi disk, can be antoher leter, like /dev/sdc, > for example). > 5 - Make two partitions for the raw shared disks. Use fdisk /dev/sda, > or parted /dev/sda, as you prefer. You must create 2 small partitions > (10 to 30 mb for each is sufficient). > 6 - Make the partitions that you need for the aplication/service data, > and format it with mke2fs -j /dev/sdaX (X is the partition number). > 7 - Configure the raw partitions, adding in /etc/sysconfig/rawdevice > two lines like: > /dev/raw/raw1 /dev/sda1 > /dev/raw/raw2 /dev/sda2 > and restart rawpartitions with /etc/init.d/rawpartitions restart > 8 - Continue with the standard installation described in the redhat > cluster suite manual guide. > > I have do all this, and I must say that the cluster is working very > well. The switching speed is very fast, and the general performance is > also very good. > > I hope this guide can help somebody now or in the future ;) > > Best Regards > Marcos > > > On Wed, 26 Jan 2005 13:51:25 -0500, Lon Hohberger wrote: > > On Wed, 2005-01-26 at 14:42 +0100, Marcos Monge wrote: > > > > > There is any easy way to upgrade from update 3 to update 4? use > > > up2date to install the latest updates/erratas is sufficient to have > > > the iSCSI features? > > > > Should be. > > > > -- Lon > > > > > _______________________________________________ Piranha-list mailing list Piranha-list at redhat.com https://www.redhat.com/mailman/listinfo/piranha-list From lhh at redhat.com Fri Jan 28 20:03:12 2005 From: lhh at redhat.com (Lon Hohberger) Date: Fri, 28 Jan 2005 15:03:12 -0500 Subject: Cluster without shared disk, or over shared over NFS In-Reply-To: <6204EFD303E1D24297BBD43A6ED722E7219882@RSWMS02> References: <6204EFD303E1D24297BBD43A6ED722E7219882@RSWMS02> Message-ID: <1106942592.16910.310.camel@ayanami.boston.redhat.com> On Thu, 2005-01-27 at 15:28 -0500, Brian Beaver wrote: > Thanks for the good info. So to summarize, Redhat clustering still appears > to require "local" disks. The Netapp iSCSI solution just helps the drives > appear local to the O/S. Any idea if Redhat clustering is allowed without > the shared drive requirement, similar to Veritas clustering using multiple > NIC's for heartbeats instead of a shared drive partition? To be more precise, it's not heartbeating that needs it. You can use heartbeat over bonded ethernet links for network redundancy if desired. It's the semi-persistent shared cluster service and lock states and cluster configuration information which is stored on the shared raw partitions. The shared raw disks can be used as a backup to determine whether the other node is online in 2-node clusters in the event of a network partition. -- Lon