1 2 Previous Next 11 Replies Latest reply on Jan 15, 2010 8:05 AM by peter_eepc

    EEPC in Clustered Server Configuration

    R3k1awyk5

      Has anyone ran EEPC in a Clustered Server Configuration?  If so,  anyone have some doc that they wouldn't mind sharing?

       

      Basically, we are wanting to stand up some new EEPC servers in a HA Active\Passicve Server Cluster,  meaning if serverA goes down everything failsover to serverB.  I think this can be done by installing the EEM and EEPC apps on both servers, and setting up the database on a cluster resource that can failover between the two nodes.  Just curious if anyone has a better way of doing it, or if the know of any "gotchas"  before I get there.

       

      Thanks in advance for the help!

        • 1. Re: EEPC in Clustered Server Configuration

          I've done something like it, and it's not the easiest thing in the world, but it did save my bacon once. (no other outages since then!) <cross my fingers>

           

          Setup: Two Windows 2003 servers, same domain, different locations.

           

          EEM (was SafeBoot) installed on both servers.

          #1: Has database running as service, http server, and Object Backup running.

           

          #2: Has database installed-not running, no http server, and no backup. Has \sbadmin\SBDATA\ shared.

           

          Every weekend, the #1 server uses Object Backup to backup the whole database to the share on #2. Takes about 30 hours (24,000 machines).

          Every evening, the #1 server uses Object Backup for incremental backup to the share on #2. Takes about 3-4 hours.

           

          The installer lists both servers, both IPs, but always try the #1 server first.

          Failover is:

          Turn on the database service on #2. Make absolutely sure the database on #1 stays off until we're ready to bring it up.

           

          Return to home is:

          Stop (and disable) the DB on #2.

          Copy all the data to #1.

          Start the DB on #1.

          Test a client synch.

           

          Issues:

          Backup time is a HUGE problem. It takes a really, really long time. I suspect there's something wrong somewhere, but I haven't been able to put my finger on it. Could be server, timeouts, network, etc. I suspect that if we shut the db down before the backup, it would go a LOT faster.  It's on my todo list.

          • 2. Re: EEPC in Clustered Server Configuration

            if you are using a HA cluster, then you don't really have to do anything - just install everything into the shared storage and set the services to be HA (with failover).

             

            the cluster service should take care of the rest for you.

             

            Make sure you give it solid DAS though, or Tier I SAN if you must.

             

            Simon.

            • 3. Re: EEPC in Clustered Server Configuration

              That's almost a cold standby (from application point of view). Typical backup solution.

               

              Cluster is doable for sure. Make sure you have top notch SAN, if database is going to be large (20K+ objects).

              Please send me PM if details are needed.

              • 4. Re: EEPC in Clustered Server Configuration

                cold? No, I disagree - the 2nd+ nodes are already up and cooking, yes, it has to start the service, but the data is online already, and the network traffic is automatically passed over.

                 

                once the cluster notices a node has gone offline, it should be microseconds before another node takes over. That's not cold at all?

                • 5. Re: EEPC in Clustered Server Configuration

                  It would be true if all object data would transfer instantenously. Unfortunately this is not the case with large databases.

                  Performing EE real time backups during prime business time, always creates object/attribute misalignments (resulting in orphan objects and other problems).

                  It is not a real time (hot) backup.

                   

                   

                  I was refering to rbarstow solution. on 1/14/10 1:23:06 PM EST
                  • 6. Re: EEPC in Clustered Server Configuration

                    To clarify, in a HA Cluster there is only one copy of the DB. It's transferred instantaneously to the node which picks up responsibility.

                     

                    You don't need to do any backups or replication, control of the disk is bounced to the node by virtue of the cluster-aware OS - every node has the data available all the time, but only the active node has control.

                     

                    Of course, in an Active Active cluster, all the nodes have full access to the data simultaneously (via arbitration), but I'm not sure that's supported under Microsoft OS's yet.

                    • 7. Re: EEPC in Clustered Server Configuration

                      You keep talking about Cluster while I posted my comments on rbarstow reply, where he talks about backup.

                      By the way, is there any McAfee EE document outlining Cluster configuration in detail?

                       

                      The only reference I found is in Best Practices Guide and KB article,

                      https://kc.mcafee.com/resources/sites/MCAFEE/content/live/PRODUCT_DOCUMENTATION/ 21000/PD21801/en_US/Endpoint%20Encryption%20Enterprise%20Best%20Practices%20Guid e%20v2.pdf

                      but is not very encouraging:

                      Clustering

                      Fully automated failovers for applications usually employ a cluster server environment. Although the McAfee

                      Endpoint Encryption Object Directory and Manager can run on a cluster, we recommend against using ‘shared’

                      resources where possible. As per McAfee KB53698, Windows Cluster environment has not been fully tested at

                      this time in engineering.

                       

                       

                      on 1/14/10 4:46:22 PM EST
                      • 8. Re: EEPC in Clustered Server Configuration

                        I replied to you Peter in time order, prior to you editing your post and saying you were not replying to me :-)

                         

                        As for clusters, no, I'm not sure why we would have any documentation on it - we don't do anything specific (or need to) with a cluster - it's the job of the cluster to work it out.

                         

                        You can install anything on a cluster that you want, even calc... How to do that is all part of the cluster documentation.

                         

                        I've not done it for a few years, but apart from getting the cluster set up, it was childsplay to install EEM on it. I remember spending the whole day pulling the plug out of one Compaq box or another, they survived my torture admirably, and the database never went offline for more than a second or so.

                         

                        Simon.

                        • 9. Re: EEPC in Clustered Server Configuration
                          I've not done it for a few years, but apart from getting the cluster set up, it was childsplay to install EEM on it. I remember spending the whole day pulling the plug out of one Compaq box or another, they survived my torture admirably, and the database never went offline for more than a second or so.

                          It is quite a bit more than that, especially in bigger database environment. You have to configure number of EE specific parameters, to accomplish smooth failover, which would not require maintenance procedures after the switch. And it takes more than a second, at least by design.

                           

                          As for the replies, there is a place in reply, which clearly indicates to whom the reply was sent to (look for "in response to:"). As far as I know, this info is not editable.

                           

                           

                          on 1/14/10 8:01:54 PM EST
                          1 2 Previous Next