DSC with 20K Nodes >:}

Welcome Forums DSC (Desired State Configuration) DSC with 20K Nodes >:}

Viewing 7 reply threads
  • Author
    Posts
    • #201099
      Participant
      Topics: 7
      Replies: 12
      Points: 89
      Rank: Member

      I have been looking into DSC to assist in management of our end nodes.  There is a lot of information out there that indicates DSC Pull server is limited in management numbers, compared to Puppet / Chef / Ansible.  It appears that Puppet / Chef / Ansible are more robust and scalable.  The recommended number of nodes connected to a DSC Pull server is 1000:1.  These numbers don’t scale well.

      Does anyone have any suggestions, insight or horror stories in trying to manage a large volume of nodes?

       

      Kind Regards!

      Alan

    • #201138
      Participant
      Topics: 5
      Replies: 7
      Points: 42
      Rank: Member

      Writing our horror story.  We’re using a DSC pull server with a SQL backend. We have around 4000 nodes. When we implemented our pull server, we were unaware of this 1000:1 limit. Last September, we suddenly started having intermittent failures when servers were running consistency checks. After 4 months of talks with MS support and countless logs sent, they told us of this limit and that the recommendation is to use Azure DSC for our size environment, but they have no migration documentation available for us to migrate our current environment to Azure DSC.

    • #201204
      Participant
      Topics: 0
      Replies: 3
      Points: 15
      Rank: Member

      I don’t know anything about the technical details of that limit, but what about setting up multiple Pull Servers with the same config and put a Loadbalancer in front of it?

    • #201306
      Participant
      Topics: 7
      Replies: 12
      Points: 89
      Rank: Member

      I don’t know anything about the technical details of that limit, but what about setting up multiple Pull Servers with the same config and put a Loadbalancer in front of it?

      The issue with multiple pull servers in a pool is still the load.  I’m sure that they would be more efficient in a pool, but the issue is still the same.  In my case I would “guess” I would still have to have upwards of 10-15 pull servers.  As Eric has indicated, the MS solution is to use Azure DSC but I don’t know what that does differently to handle the high load.  It’s still reporting and pulling to/from a single source???  My guess is that still has a limit to it as well.

       

      Thanks for the input.

    • #201309
      Participant
      Topics: 7
      Replies: 12
      Points: 89
      Rank: Member

      Writing our horror story. We’re using a DSC pull server with a SQL backend. We have around 4000 nodes. When we implemented our pull server, we were unaware of this 1000:1 limit. Last September, we suddenly started having intermittent failures when servers were running consistency checks. After 4 months of talks with MS support and countless logs sent, they told us of this limit and that the recommendation is to use Azure DSC for our size environment, but they have no migration documentation available for us to migrate our current environment to Azure DSC.

      Thanks for sharing your story.  I don’t feel so bad now asking such an open ended question!  😉  I have found the reference to Azure DSC also, but I still don’t know how much of a difference it would make.  We like to keep things internal to our network as we are an K-12 EDU, but we have also been pushed by Microsoft to Azure and AWS for other things as well.  You know what they say about “cloud computing” right?  It’s just someone else’s computer(s) and network!

    • #201525
      Participant
      Topics: 0
      Replies: 3
      Points: 15
      Rank: Member

      In my case I would “guess” I would still have to have upwards of 10-15 pull servers.

      You’re managing 20K nodes :-). I wouldn’t take care about 20 more.

      Maybe setting up the refresh rate up to >180 minutes will reduce the load.

      Do you know the bottleneck of a DSC Pull Server -> CPU, Memory, Storage? Or is the poor performance just by design

    • #201570
      Participant
      Topics: 7
      Replies: 12
      Points: 89
      Rank: Member

      In my case I would “guess” I would still have to have upwards of 10-15 pull servers.

      You’re managing 20K nodes :-). I wouldn’t take care about 20 more.

      Maybe setting up the refresh rate up to >180 minutes will reduce the load.

      Do you know the bottleneck of a DSC Pull Server -> CPU, Memory, Storage? Or is the poor performance just by design

      To my knowledge, it’s poor performance by design…  🙁  I wonder though, going back you your original comment, if a DSC pool to a singe SQL server could manage a higher load, but I have no idea what the ration would be…

    • #201618
      Participant
      Topics: 0
      Replies: 3
      Points: 15
      Rank: Member

      The performance should scale pretty well. IMHO the pull server is just a stupid webserver which provides some mof files to download (+some reporting mechanism). I am neither an dsc nor an sql expert, but a well sized SQL shouldn’t be the bottleneck.

      I would give it a try.

Viewing 7 reply threads
  • You must be logged in to reply to this topic.