Skip to end of metadata
Go to start of metadata

The following does not apply to CoprHD deployments or CoprHD devkits.

 

As part of continuous improvement, one of the major productivity issues we’ve dealt with over the months has been issues with providers getting bombarded with discovery, metering, and scan queries from dozens/hundreds of controller instances.

Primarily this has impacted the reliability/stability of our SMI-S providers, but also has far-reaching effects on our switch providers, VPLEX, hosts, and RP providers as well.

In an effort to reduce the strain on these resources and regain some stability, I’ve disabled by default any single node deployment’s auto-discovery feature starting in X-wing (master).

You are welcome to turn it back on by setting prop “enable-auto-discovery-metering-scan-single-node-deployments” in controller-configinfo.xml or through syssvc and restart the services.

This impacts devkits, Standalone deployments, and 1+0 deployments.  This does NOT impact 3 or 5 node deployments.

This also impacts those who use single node deployments to test things like “Capacity Verification” (where you like to see the capacity numbers change in Controller on the hour when stuff happens on the array underneath controller) and also “Port Metrics” where you like to see controller make different placement decisions based on port traffic.  Usually automated tests in those areas require manual discovery, so it shouldn’t be an issue, but I’m required by law to warn you. 

We have further improvements we would like to make in the area of discovery to reduce impact, but this should go a long way in removing the noise associated with down or over-stressed providers.  Not to mention it makes our logs a lot easier to read.

Developers: As an added bonus, I pulled sys-metadata-var.xml and sys-metadata-var-template.xml from source code control, since they are generated as part of the build.  Hopefully you don’t have to deal with unchecking-out those files anymore, but if you see any issues with me pulling it, such as using the product in “runtime” mode on your local desktop, please let me know!

https://coprhd.atlassian.net/browse/COP-22155

https://review.coprhd.org/projects/CH/repos/coprhd-controller/pull-requests/2577/overview

  • No labels

2 Comments

  1. Bill Elliott, given that most CoprHD deployments are single node, does this mean that pretty much everyone using CoprHD needs to turn this back on if they want the system to actually discover their storage?  Or will it still perform the initial discovery but just not update on a periodic basis?

     

    Also, you talk about turning it back on in syssvc. Is this going to show up in the UI's General Configuration→Controller area?

     

    Thanks!

    1. If this is enabled, you still get initial discovery when you register the storage so your ports and pools, etc, are all filled-in.

       

      I'm looking to remove this for CoprHD anyway due to a recent defect, and because CoprHD is typically a single node deployment and turning this off may not be a great idea from that perspective.  I'll update the wiki appropriately.

       

      No, I did not plan on putting the variable in the UI.  If there's a push, I can add it.