Skip to main content

Very slow adding servers



  • Sam Kirchoff

    This likely a good blog entry too!

    By design DPACK or the future Live Optics is designed to assist in project level data gathering that is needed for decision making. This would be in contrast to entire Datacenter scans.

    So the first question is simply "are all 600 servers in a current IT project that is being evaluated?"

    If not, and you just want to use the collector as an inventory utility, then we suggest you limit the number of servers per instance to 100-200 servers. However, it is fully supported to spawn multiple instances of the collector and add them to the same project for aggregation.

    Last, if these contain a high percentage of virtual machines, then we suggest you collect through the Hosts alone. If you collect from the Hypervisors perspective it will catalogue the VMs for you automatically.

    However, if you collect the Hosts and all the VMs as targets together, then you are effectively double counting all your performance needs. The aggregate performance demand from the Hosts will be similar if not near equal to the aggregate view of all the VMs.  So you want to choose one technique or the other.

    Please feel free to continue this dialogue. As with lots of things, it's often more about technique than capability.

  • Tom McElvogue

    Thanks for the information Sam. The reason we are trying to scan the entire datacentre is that we are looking at the feasibility/sizing of a hyperconverged architecture and for that we were wanting to get a broad picture of the entire estate. Perhaps we need to come at this from a different angle but I understood that this tool would accomplish that task.

    I was aware that the aggregation was possible afterwards however from what I read, the scans had to all commence within the same hour in order for this to work (which would make sense). Unfortunately, adding in 20 servers can take an hour given that each connection is made right at the start before launching a "collector" scan. The software has crashed several times without warning during this setup process which has caused considerable delay and frustration.

    We have 300+ Xenservers which don't seem to be supported and that forms quite a large part of our analysis group which also causes this exercise to be quite problematic.

    I wasn't aware that the Host would automatically pull in the VM's using HyperV and providing this gives a level of granularity that we can then further explore via the reports that could work.

    Thanks again for the response.

  • Sam Kirchoff

    Xen is definitely supported. There might be more to this as each environment is unique. We can assign a support team member to help you get there. however, in any circumstance, this is a huge undertaking.

    The collector can also import files to help start with similar start times. however, the one hour thing is really an older recommendation from days gone by.

    When we pull in VM data through the host it will have config detail etc... but not performance. We don't do that on performance since it would cause that double count I mentioned. Let me get someone to teach out.


Please sign in to leave a comment.