ruben@bcstec.boeing.com (Reuben Wachtfogel) (06/26/91)
I'm working on a ring of >100 HP/APOLLOs and have a single node that stores several sensitive data files that must be available for our major application to work. I could store dual copies of these files on 2 seperate nodes and code the application to perform all updates of these files on this MASTER node as well as a BACKUP MASTER node so that if the MASTER were vaporized, the BACKUP could be slipped in with minimal downtime. My question is: 'Is there some application transparent way to achieve data redundancy in a unix network ?' I would think that a DEVICE DRIVER could be written to accomplish this. Is there an elegant way ? SYS5 Streams ? Does NFS help ? Well Wizards, what say You ? -------------------------------------------------------------------- Please respond here or to ruben@dsp35001.boeing.com Thanks If you don't like the topic, ignore it or post your flames to /dev/null. DISCLAIMER: This posting was randomly generated by chimp #2306893743823 and does not represent the opinion of any large Aerospace company starting with a 'B'.
jasonp@cunix7.prime.com (Jason Pascucci) (06/26/91)
In article <992@bcstec.boeing.com>, ruben@bcstec.boeing.com (Reuben Wachtfogel) writes: |> I could store dual copies of these files on 2 seperate nodes and |> code the application to perform all updates of |> these files on this MASTER node as well as a BACKUP MASTER node |> so that if the MASTER were vaporized, the BACKUP could be slipped |> in with minimal downtime. |> |> My question is: |> |> 'Is there some application transparent way to achieve |> data redundancy in a unix network ?' |> I haven't heard about anything which would do this for you, but here are a few ideas which might solve your problem: NFS helps a great deal. I don't know if your current filesystem scheme will do this, but when you make an NFS request to a site which is down, it will hang, not crash. This is a big plus from your point of view. (I personally hate this, esp. with my Sun. Ick) Now, there are a few things you could do. How about setting up a 'Secondary' system which has a process that constantly checks to make sure the 'Master' is alive. If it isn't, have it restart TCP/IP with the address of the 'Master', and NFS will resume talking with you instead of him. When the Master boots back up, have it copy over the current data, and Re-Shutdown the Slave, and have the Master take over yet again. Admittedly, it's not elegant, but it should do what you want with a minimum of work. If you want to throw a little hardware at this, You could combine the Network stuff with Dual Ported SCSI, assuming the OS supports Priority Select. This would allow you to use extremly up-to-date data, and is a large win. This doesn't guard against disk failures, though. Another solution is to have backups on the 'Secondary' system, and still do the dual porting. You may also want to look into Third party vendor (I don't think HP's unixii do this, but I don't claim to know for sure) who do disk mirroring, and Dual port one half (Or both) of the mirror. There are all sorts of Hardware you can throw at the problem, if you want to spend the money. Your Milage May vary. These are, of course, only suggestions. -- Jason R. Pascucci "Kate Bush Is God!......Oops. Wrong newsgroup" jasonp@primerd.prime.com Disclaimer: My company isn't responsible.
petri@ibr.cs.tu-bs.de (Stefan Petri) (06/28/91)
In article <1991Jun26.112529@cunix7.prime.com> jasonp@cunix7.prime.com (Jason Pascucci) writes: >In article <992@bcstec.boeing.com>, ruben@bcstec.boeing.com (Reuben >Wachtfogel) writes: >|> I could store dual copies of these files on 2 seperate nodes and >|> code the application to perform all updates of >|> these files on this MASTER node as well as a BACKUP MASTER node >|> so that if the MASTER were vaporized, the BACKUP could be slipped >|> in with minimal downtime. >|> >|> My question is: >|> >|> 'Is there some application transparent way to achieve >|> data redundancy in a unix network ?' >|> > >I haven't heard about anything which would do this for you, But I have : What you want is to have a look at rdist(1) rdist - remote file distribution program To keep the contents of your Backup consistent with the Master, and at amd(8) to create replicated NFS-Servers for your Users/Clients amd - automatically mount file systems rdist should be contained in any reasonable version of Unix, maybe its also in the freed parts of the BSD-Sorces (I didnt check that). Amd is the BSD4.4-Automounter, but it runs on a broad variety of Unix'es It's operation principles are similar to Sun's automount, but is has more capabilities, especially to configure Backup Servers that get automagically used when the Master is down. Rdist and amd together work like a charm for us since several month now. amd is available from usc.edu in ~ftp/pub/amd . S.P.