Topic
4 replies Latest Post - ‏2013-09-13T16:42:06Z by bwchang
Ziliotti
Ziliotti
5 Posts
ACCEPTED ANSWER

Pinned topic Configure Omnifind Enterprise Edition 9.1 in HA (Active-Passive) via HACMP

‏2012-08-01T14:53:38Z |
Hello,

I have some doubts about configure the Omnifind EE 9.1 in Active-Passive cluster. Here is my cenario:

We have 2 IBM Power servers, each LPAR is dedicated to Omnifind installation. In the "server A" we´ve installed the Omnfind EE 9.1 with the config: "All-in-one(Master)Server", it´s running fine and done.
When the "server A" down, we would like which the HACMP move to "server B" all resource groups and configurations from Omnifind, and start the server normally (we have a VIrtual IP configurate too).

My doubts are:
1-) In the "server B", we´ll install the Omnifind binarie with the "All-in-one(Master) Server" configuration or other ?
2-) In HACMP, we´ll configure to move resource, only the ES_NODE_ROOT directory ? Or we have to configure other directories and files?
3-) Are there other ways to install and configure the Omnfind in this kind of topology/cluster (Active/Passive)?

e.g.: I read and understand the High availability support in this new version to Omnfind, but we have a problem about the license to configure a SAN/NAS shared file repository. In this momento, our option to solution is use the Active/Passive (move resources) from HACMP.
Thanks a lot for time and help.

--
Sergio Ziliotti.
Updated on 2012-08-27T14:27:00Z at 2012-08-27T14:27:00Z by Ziliotti
  • bwchang
    bwchang
    146 Posts
    ACCEPTED ANSWER

    Re: Configure Omnifind Enterprise Edition 9.1 in HA (Active-Passive) via HACMP

    ‏2012-08-24T20:42:48Z  in response to Ziliotti
    In general, there are two types of High Availability within OEE 9.1.

    1) Search HA, only search is configured for High Availability
    2) Full HA, the whole system is configured for High Availability (may experience 15-30min down time when overtake is in process)

    To achieve Search HA, you simply needs to configure OEE with at least 2 servers that has search capability. For example, Server A can be configured as All-in-one Master server that has everything running there, including search. Then Server B can be configured as an additional Search Server, to Server A. Then you need to front these two server with a network dispatcher server (a.k.a. HTTP load balancer, sprayer, or router), that will route all search requests to either Server A or Server B depending on the load of the servers. In this case, both servers are active.

    To achieve Full HA, you need to configure Server A as the All-in-one Master server and Server B as the All-in-one High Availability server (or Master Backup server). Please see this link that documents additional server. In this case, the additional server will be the High Availability server. In this case, Master server is active while the High Availability server is passive.

    http://publib.boulder.ibm.com/infocenter/discover/v9r1m0/index.jsp?topic=%2Fcom.ibm.discovery.es.in.doc%2Fiiysiservrenhadd.htm

    Once both Master server and High Availability server is installed, you need to configure IBM PowerHA for both servers. Please see instruction for this on the link below.

    http://publib.boulder.ibm.com/infocenter/discover/v9r1m0/index.jsp?topic=%2Fcom.ibm.discovery.es.in.doc%2Fiiysihaaix.htm

    With that said, we also noticed some customers roll out their own HA solution.

    1) Install 2 independent All-in-one Master servers that are not aware of each other. In this case, the customer will create a collection and its crawlers on Server A and run the export / import of a collection to Server B, to keep configurations in sync manually. While this is a simple set up, there are a lot of manual efforts and both needs to crawl, parse, and index all documents. This is active/active.

    2) Install 2 independent All-in-one Master servers where Server A is active, and Server B is passive. If Server A goes down, Server B would perform a complete takeover, by

    a) shutdown Server A
    b) hijacking the file system (both ES_INSTALL_ROOT and ES_NODE_ROOT), IP and hostname
    c) startup Server B

    Of course, a lot of custom set up is required to perform automatic switch over. I do not have any detail on this implementation, but was told it was done successfully.

    Hope this helped.

    Billy
    • Ziliotti
      Ziliotti
      5 Posts
      ACCEPTED ANSWER

      Re: Configure Omnifind Enterprise Edition 9.1 in HA (Active-Passive) via HACMP

      ‏2012-08-27T14:27:00Z  in response to bwchang
      Hi Billy,

      Thanks for answer. I gonna make the configurations and procedures who you describe and after I´ll post the result.
      Thank you for answer.
    • iamjry
      iamjry
      1 Post
      ACCEPTED ANSWER

      Re: Configure Omnifind Enterprise Edition 9.1 in HA (Active-Passive) via HACMP

      ‏2013-09-13T01:50:46Z  in response to bwchang

      Hi Billy,

      Thanks for your post. I got two questions for you,

      1. I only saw All-in-one server, didn't see All-in-one "High Availability" server , is that the same ?

      2. Should I make the folder ES_NODE_ROOT accessible for two servers in the same time , if I want to config a Active/passive mode?

      Thanks.

      Regards, Jerry

       

      • bwchang
        bwchang
        146 Posts
        ACCEPTED ANSWER

        Re: Configure Omnifind Enterprise Edition 9.1 in HA (Active-Passive) via HACMP

        ‏2013-09-13T16:42:06Z  in response to iamjry

        Hi "iamjry",

        1. From the Information Center, the passive backup server is called "High Availability" server.  When adding this server to the Master Server from ESAdmin administrative console, the Server role is called "Backup index, crawler, and search server".  Note that this is supported only for Windows and AIX platform.  For Windows, we are leveraging MS Clustering Service.  For AIX, we are leveraging PowerHA.  Both have specific configuration instructions.

        2. For both the Master server and the Backup server, then should have ES_NODE_ROOT configured with the same path, from the local disk.  However, this does not imply shared disk where both servers access the same folder via mount.  Mounting will create another single point of failure, which is something we are trying to avoid for high availability.

        Hope this is what you are looking for.  If not, please ask again.

        Thanks, Billy.