Results 1 to 3 of 3

Thread: Hana and SAP on the same node, 2 Nodes configuration for HA

Threaded View

  1. Re: Hana and SAP on the same node, 2 Nodes configuration for

    Quote Originally Posted by ch061441 View Post
    Dear Comunity,
    we have following Problem, we have a two nodes configuration,
    we have SAP Hana running with SAP Hana Replication between the two nodes
    and SAP ASCS and PAS installed on the first node (No ERS)

    We can move SAP from one nodeA to nodeB without any problems, this works fine, and also HANA Take over works fine.

    Our problem is following:
    When we put the Primary Node in Standby first of all the PAS and ASCS should be stopped and then the HANA take over should be started and in our case all runs in parallel,
    stop of PAS and take over.
    We need some constraint, colocation, tried some ones but without success.
    Does any one of you the same configuration, or one similar?
    or some examples of a configuration where SAP is installed on the same node with HANA?
    or just some recommendations how to configure this constraint?

    Regards.

    Manuel
    You will need an order constraint.
    The help is viewable via:
    Code:
    crm configure  order --help
    In your case the orders should be something like this:
    Code:
    crm configure order start-HANA-before-PAS Mandatory: SAPHanaController:promote then PAS:start
    crm configure order stop-PAS-before-HANA Mandatory: PAS:stop then SAPHanaController:demote
    I assume you are using 2-node + MajorityMaker. (If not -> you are in deep trouble !!!)
    If that's true, you can consider to keep both Master resource SAPHanaController and the PAS together.
    Then you will need a collocation rule like this one.
    Code:
    crm configure colocation PAS-with-HANA inf: PAS:start SAPHanaController:promote
    Note sometimes failiure of only 1 resource is possible (loss of shared storage for PAS, crash of master HANA,etc), so try as many failiure scenarious before putting the cluster in production ! Setting a node into standby is not enough (it's OK to test ordering), so either invoke a crash of one of the cluster components, block communication via the firewall, or close the network ports to one of the nodes.
    It's always better to find a flaw and correct it before the system is in production, than to fight design issues later.

    P.S: I hope you have checked the following article
    Last edited by strahil-nikolov-dxc; 23-Jan-2020 at 19:41.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •