Announcement

Collapse
No announcement yet.

Redundancy - failover OS?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Redundancy - failover OS?

    Any thoughts re auto failover techniques guys?

    I have a feeling it's one of the options with VMware. I'm thinking along the lines of two OSs running in VMware, primary and secondary so to speak, such that if OS server 1 fails OS 2 detects this and automatically becomes responsive and begins to serve. OS 2 then reports this to admin via email.

    Any thoughts?

    BTW I have only one machine so VPSs appears the most natural route.


    Cheers
    Lea

  • #2
    What OS?

    I must say it seems a little pointless as there are still so many SPOFs left...

    I'm guessing here but...

    Single Machine
    Single Net Connection?
    Single Power Supply?
    Single HDD?

    Any one of those fail and whats the point in having failover VMWare images? Add to that that VMWare itself could crash.
    Goscomb Technologies Limited - /
    E: P: +44 (0) 203 129 4400 F: +44 (0) 203 129 4410 T:

    !

    Registered in England and Wales No. 05672987 - VAT Registration No. 853 7954 80

    Comment


    • #3
      ... and not to mention the Host OS
      A very welcome new user!
      Sig is gone! Please read the sig rules before applying one!

      Comment


      • #4
        W2K3 Standard

        I take your point re hardware, however in my limited experience the software system as a whole (IIS, OS, etc) has a significantly higher risk of failure. So, if one were to exclude hardware issues then software redundancy should be a priority.

        Or maybe it doesn't and the data I'm basing my conclusions is corrupt or inaccurate. So, I should ask - in your experience what's the most common cause of loss of service?


        Cheers
        Lea

        Comment


        • #5
          In our experience, hard drive has been the thing that we've had to replace the most.

          Other things closely following that, are customers messing up their installs/installing software that kills the OS (eats too many resources, etc)..
          A very welcome new user!
          Sig is gone! Please read the sig rules before applying one!

          Comment


          • #6
            Originally posted by LeaUK2 View Post
            W2K3 Standard

            I take your point re hardware, however in my limited experience the software system as a whole (IIS, OS, etc) has a significantly higher risk of failure. So, if one were to exclude hardware issues then software redundancy should be a priority.
            Right... so you still have your OS as a SPOF if thats your worry... as VMWare runs on top of an OS (the affordable versions anyhow)
            Goscomb Technologies Limited - /
            E: P: +44 (0) 203 129 4400 F: +44 (0) 203 129 4410 T:

            !

            Registered in England and Wales No. 05672987 - VAT Registration No. 853 7954 80

            Comment


            • #7
              The question is what are you trying to acheive ?
              If its a critical application/service then why are you running it on a VPS which introduces more points of failure, rather than load balanced dedicated servers or whatever ?

              What "problem" are you tryingto solve ?
              Rob Golding Astutium Ltd AS#29527 Company#08183381 Phone#020 3475 2555
              Domain Name Registration - uk domains just £5.55/2 years | DNS Services | Web Hosting from £2.95 | Minecraft Servers from £2.50
              London Docklands Colocation 1u £49.95 | Virtual Private Servers from £4 | Virtual Dedicated Servers from £12 | Managed + Unmanaged Dedicated Servers from £69
              Make more money from domains - Talk to me about our Domain Reseller Accounts and WHMCS modules

              Comment


              • #8
                if your running the free version of VMWare it still needs to be installed on an operating system.

                OS - VMWare - 2 x OS

                If you were wanting to run your 2 VPS as a windows failover cluster you would have to run W2K3 Enterprise as clustering isnt included in teh standard or web versions, this would be rather costly. You could still run your base OS as linux to save some cost there.

                As i'm just in the middle of implemeting a VMWare DR system i have a little knowledge on this subject, so feel free to ask.

                Modern operating systems are alot more stable than you seem to think and i really do feel that your over complicating things here.

                KISS
                Martin Robinson
                (url-removed: need 20 posts)
                Sales: 0845 838 2379

                Comment


                • #9
                  O/S Failure? I'm not sure i've experienced that too much..... I've heard of hacked boxes and i've heard of people breaking boxes with badly managed updates etc...

                  If you leave a machine running and never touch it there is a good chance it will sit there forever until some hardware borks.
                  Mooharr
                  E-Mail Hosting Services

                  These are not my views and i cannot be held accountable for anything he says.

                  Comment


                  • #10
                    I have to say that we run Virtuozzo on 95%+ of our servers and very very rarely have system failures at the host OS level, so this would increase redundancy in many situations. However, I agree that it's probably not wise having 2 VM/VPS's on the same host OS if the aim is to improve redundancy, unless your applications are known to crash frequently. I have no idea how VMware compares to VZ with stability at the host OS level.
                    Andrew Cranson :: Director of Operations

                    Layershift :: UK, New York, Singapore :: Managed Cloud VPS, Dedicated, and Complex Hosting
                    Find out what fully managed hosting really means!

                    Company Number: 6036217 / VAT Number: GB903034962

                    Comment


                    • #11
                      Many thanks for all your replies, much appreciated.

                      My concerns are based on my limited experience of setting up the base OS and running a web server for a month. For example, I had real problems upgrading mySQL 4 to 5 which ended up requiring a W2K3 reinstall, and that would have been a pain had the server been externally located - although with a remote KVM and a set of HDD images I think I could handled it. And yesterday the server was simply not responding, I made no changes - just dead for no appernt reason. A reboot sorted it but how long was it down? Could have been all night!?

                      are customers messing up their installs/installing software that kills the OS (eats too many resources, etc)..
                      This worries me too... but I'm not sure if VPS are the correct solution here either...?

                      Your general comments re OS stability (even Windoz) has changed my perception and as most say, maybe stability is more a function of HDD MTBF nowadays. HDD failure is also my opinion of the number one failure of hardware within PCs!

                      I think the comments about layering OSs with VPSs are also probably true - maybe I'll be fine with simply a remote KVM and several OS images?

                      My requirement for redundancy was simply to allow contiguous service, I don't mean 99.99999999% nonsense as my server is only for friends/family at this stage, but was hoping for >95%.... do you think this level is achievable?

                      I take all your comments all on board and like to say thanks for your input as it's you experienced guys that can offer such great practical advice.

                      Cheers
                      Lea
                      Last edited by LeaUK; 11th November 2006, 12:28 PM.

                      Comment


                      • #12
                        I wouldn't suggest you set up a VPS in this situation.
                        Last edited by folklore; 11th November 2006, 01:48 PM.

                        Comment


                        • #13
                          Originally posted by LeaUK2 View Post
                          And yesterday the server was simply not responding, I made no changes - just dead for no appernt reason.
                          That's a little concerning. Nothing in the logs? I would keep the machine running under load for a while before you take it to the DC, to see if it happens again. Maybe do some hardware testing too (memtest etc).

                          Originally posted by LeaUK2 View Post
                          A reboot sorted it but how long was it down? Could
                          have been all night!?
                          You should probably set up some remote monitoring to call you when the server stops responding. Also make sure you have the ability to power cycle the machine remotely when it is at the DC.

                          Originally posted by LeaUK2 View Post
                          My requirement for redundancy was simply to allow
                          contiguous service, I don't mean 99.99999999% nonsense as my server is
                          only for friends/family at this stage, but was hoping for >95%.... do you think this level is achievable?
                          So long as your hardware is good and you have no issues keeping the software updated, it should be achievable.
                          Last edited by psi; 11th November 2006, 01:58 PM.

                          Comment


                          • #14
                            Tried using Win NLB on two boxes

                            Works no issues on our NLB Cluster, for example for some unknown reason the NLB nic on one of our boxes stopped accepting connections nlb took over automatically and put full 100% load onto the working box.

                            There’s only one problem for example if IIS starts being a stupid son of a gun i.e. crashes then nlb cant tell if its running or not so it still sends requests to that box.

                            Now our solution to the above problems is using siteuptime.com when it detects a failure on IIS it sends an email the mail server parses it to see which box and if its IIS then it issues an nlb command to stop issuing IIS requests to the box which isn’t working

                            I’m working on the above solution at the moment, our mail server software Merak is quite powerful so its defo possible for us.

                            So far the NLB has worked perfectly apart from the IIS issue above, at least when the nic played up on the second box NLB detected it within 30secs and sorted its self out

                            The great thing though with our setup, if you need to upgrade mail on one box you can disable port 25, 110 through nlb manager for the machine in question, complete your mail upgrade, then hit the enable button. Same goes for MySQL we have MySQL 5 in master to master replication mode so we can upgrade one box then the second box without any downtime to clients.

                            Anyways good luck
                            Chris Daley
                            DWebs Ltd :: Company No. 05603664 :: 0330 22 90 666
                            DHosting.uk - Shared/Reseller Hosting, Dedicated Servers, Managed Hosting Solutions

                            My views are my own and not those of my company.

                            Comment


                            • #15
                              siteuptime.com great tip thanks.

                              NLBs - way to complex for me at this time...but fascinating nether less.

                              Big thanks for your input and advice.

                              Lea
                              Last edited by LeaUK; 12th November 2006, 11:02 AM.

                              Comment

                              Working...
                              X