1 Reply Latest reply on Dec 6, 2012 1:10 PM by Jörg Hoh

    Is it possible to use shared storage on the publisher and dispatcher nodes?

    skeedune

      When going through the sizing documentation for our website expansion it looks like the publisher directory will be around 1.5 TB of storage and the dispatcher will use around 600 GB of storage. We have multiple publishers and dispatchers (using apache) so the amount of storage is growing considerably - somewhere around 20 TB just for production. It is possible to use veritas filesystem clustering in order to just have one 1.5 TB share for publishers and one 600 GB share for dispatchers? Has anyone run in a configuration like this? Are there any downsides?

        • 1. Re: Is it possible to use shared storage on the publisher and dispatcher nodes?
          Jörg Hoh Adobe Employee

          Each CQ5 instance is directly connected to a repository. You cannot run multiple CQ5 instances on top of a single (physical) CRX repository.

           

          But there is room for optimization, as you can share the datastore, which usually holds the largest amount of data of a CRX repository (mostly > 80%). I don't know Veritas FS, but it should sufficient for handling the datastore requirements. The only drawback you get is that the datastore garbage collection gets a bit more complex (but it's documented). Look at http://helpx.adobe.com/crx/kb/HowToCombineTheDatastoreToPreserveDiskSpace.html for a documentation of this scenario.

           

          Regarding dispatcher caches: Theoretically sharing the filesystem with the cache files is possible, but I've heard some rumors about locking issues on FS level (IIRC they used NFS). So if you use Veritas, you might be successfull. Just give it a try.

           

          Jörg