We're using Samba, and it's working very well, however I'm interested in finding out what approaches would work for having some sort of redundancies in place in case the main server crashes. Is there anything in place for Samba in order to have some sort of backup in place in case the main server crashes. The reason is that the server is critical to the company's business, and the few outages we've had (due to reasons other than Samba) have shown us the need for some redundancy. If anyone has any information or any ideas on how to go about this, please reply. Thanks, Ben
Benjamin Suto wrote: | We're using Samba, and it's working very well, however I'm interested | in finding out what approaches would work for having some sort of | redundancies in place in case the main server crashes. Is there | anything in place for Samba in order to have some sort of backup in | place in case the main server crashes. If you're doing a full-fledged "bring up new server with same ip address and disks" kind of failover, samba will work automagically. At the opposite extreme, you can have a commented-out section in the smb.conf file that brings up a virtual server for another department, and put it in place manually when the other department's server blows up. In between is in between (:-)) If this is on a Sun, sen me mail and I'll ask a real clusters Guru... --dave -- David Collier-Brown, | Always do right. This will gratify some people 185 Ellerslie Ave., | and astonish the rest. -- Mark Twain Willowdale, Ontario | http://java.science.yorku.ca/~davecb Work: (905) 415-2849 Home: (416) 223-8968 Email: davecb@canada.sun.com
Here at ITT Avionics, we have Samba installed on a Dec (now Compaq) TruCluster. It works great! Better than an NT cluster, in many respects. There are a couple of tricks that you need to know. Someone may want to put this in a FAQ somewhere or some document, as it might be useful to someone. 1. Use smbd -D -s /somepath/cluster.conf or nmbd -D -s /somepath/cluster.conf when you start up your smbd and nmbd daemons. Don't use the default smb.conf path that is compiled in. 2. create that cluster.conf file to look something very much like this: # To edit the cluster services, remove and relink /usr/local/samba/lib/smb.conf # to the /somepath/netbios_name.conf file of your choice and then # use swat to reconfigure. # Need this so that Samba can authenticate with the correct domain # All of the rest of the information is in the netbios name of the # cluster service, followed by .conf workgroup = YOUR_WORKGROUP_HERE # This line will include the correct configuration file for the # particular cluster service: include = /somepath/%L.conf 3. When using SWAT, don't use it to modify your cluster.conf file, change the /usr/local/samba/lib/smb.conf file to link to whichever cluster service you want to modify. 4. Create service configuration files, one for each "virtual server" that you have. The %L in the cluster.conf will point to the Netbios name for the virtual machine. The only problem that I see with this whole setup is that locking information doesn't get transferred over to the other system when a cluster failover occurs. This is not an issue for us here, since we usually only have one user accessing a file at a time. I have been using this solution for about 2 years now, on various different versions of Samba. I am running 2.0.5a for Samba and 4.0-f and TCR1.6 right now for Compaq Tru64 Unix (formerly Dec OSF). I don't know how it would play with Solaris, HP, Linux, etc., but I would love to hear from you to find out what you decide to do. Matthew G. Selick UNIX/NT System Administrator ITT Avionics Matt@Selick.com (973) 284-4744
At 05:18 14/08/99 +1000, you wrote:>Date: Thu, 12 Aug 1999 10:56:03 -0400 (EDT) >From: Benjamin Suto <ben@amvalue.com> >To: samba@samba.org >Subject: Samba High AvailabilitySorry I'm a little late with this info. We use to Sun Sparc workstations with a dual-ported Sparc Storage Array linked between them. The software which controls who (in terms of system) has control of the disk groups is Veritas Volume Manager. High Availability is handled by RSF-1 (www.rsi.co.uk/products/rsf/) which monitors 3 independent heartbeats and will initiate failover when one system dies. It will run init.d / rc.d type scripts to start the services on the backup machine and so in theory anything which can be started in this way will be capable of being failed over to a backup system. A Linux version is available (free for non-commercial use). Mike Robinson>We're using Samba, and it's working very well, however I'm interested in >finding out what approaches would work for having some sort of >redundancies in place in case the main server crashes. Is there anything >in place for Samba in order to have some sort of backup in place in case >the main server crashes. > >The reason is that the server is critical to the company's business, and >the few outages we've had (due to reasons other than Samba) have shown us >the need for some redundancy. > >If anyone has any information or any ideas on how to go about this, please >reply. > >Thanks, > >Ben---------- This e-mail is intended for the addressee named above. As this e-mail may contain confidential or privileged information if you are not the named addressee or the person responsible for delivering the message to the named addressee please telephone us immediately. The contents should not be disclosed to any other person nor copies taken. Michael Robinson Network Administrator Direct line: +44-1376-505409 SDL Integrated Optics Limited Waterside Business Park, Eastways, Witham, ESSEX, CM8 3YQ. United Kingdom. Tel. +44-1376-502110 Fax. +44-1376-502125 ---------- You are invited to visit our web site: http://www.sdli.co.uk/ ----------
On 1999-08-13 15:10:47 +1000, Benjamin Suto wrote:> We're using Samba, and it's working very well, however I'm interested in > finding out what approaches would work for having some sort of > redundancies in place in case the main server crashes. Is there anything > in place for Samba in order to have some sort of backup in place in case > the main server crashes.We have Samba running under HP's MC/ServiceGuard. 2 Machines have physical access to the same disks. When one machine fails, the other will mount the filesystems, aquire the IP address, and start nmbd with the appropriate config file (smbd is started via inetd, and uses %L determine its "personality"). Clients will then (re-)connect to the other machine, and theoretically not detect any difference. In reality, Win95 doesn't handle the reconnection not as transparently as it should, so the clients will notice it. The good thing about this is that you always have current data, the bad thing is that you must view the 2 machines as one system and for every change one one machine think about possible consequences at the other machine. hp -- _ | Peter J. Holzer | Nobody should ever have to be |_|_) | Sysadmin WSR / Obmann LUGA | ashamed if they have a secret love | | | hjp@wsr.ac.at | for writing computer programs that __/ | http://wsrx.wsr.ac.at/~hjp/ | actually work. -- Donald E. Knuth -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 371 bytes Desc: not available Url : http://lists.samba.org/archive/samba/attachments/19990816/54863e65/attachment.bin