Results 1 to 6 of 6
  1. #1
    Join Date
    Mar 2002
    Location
    Ireland
    Posts
    181

    Unanswered: 2 box failover / clustering solution for oracle

    Hi gentlefolk,
    I was wondering if anybody has any thoughts on the following:
    I will have 2 solaris boxes available to me (E280 or E420) and I want to put in place an easy to use redundency / failover solution for my Oracle Database.
    I have Oracle Enterprise Addition 8i (8.1.7.4) - I am looking at multimaster replication at the moment but was wondering if people have some better ideas.
    Thanks for any ideas or help on the issue,
    Breen.

  2. #2
    Join Date
    Sep 2002
    Location
    Ohio
    Posts
    204
    Hi Breen,
    Unfortunately this is not an easy answer. It really depends on the business need for High Availability or Disaster Recovery, they are not the same. Mulimaster replication is one solution. Oracle Standby database is another. A third solution might be hardware clustering with shared disk. And, there's always Oracle Parallel Server.
    With the exception of OPS, we have used each of these solutions for different applications and have combined the hardware clustering with Oracle Standby database for both HA and DR.
    Each solution has different cost/benefit numbers. The answer as to which is best depends on what kind of outage your business can withstand and how much money you have to spend. OPS is great, but we have not been able to get a business area to loosen up their purse strings and pay for it.

    If you have any additional information to supply, maybe you can get a better answer. Also, I can send you a couple of white papers that compare and contrast a number of solutions.

    HTH,
    Patrick

  3. #3
    Join Date
    Mar 2002
    Location
    Ireland
    Posts
    181
    Thanks for your reply Patrick,
    At the moment we have a Sun Clustering solution but we want to move our message store onto it so I have been given two decent boxes to come up with a solution for Oracle to move to.
    The database needs to be available 24/7 so I am looking for this type of solution.
    Looking at OPS I noticed that there are a few extra pieces of software to install and I guess a review of the Oracle contract.
    How easy is it to set up and manage the different solutions you have suggested?
    Money isn't a major stumbling block - I would like your opinion on the best solution for a HA OLTP environment.

    Again thanks for your help Patrick,
    Breen.

  4. #4
    Join Date
    Sep 2002
    Location
    Ohio
    Posts
    204
    Unfortunately, there's not a straight forward answer. Yes, OPS is difficult to set-up and manage, but only until you get it figured out. With OPS and Transparant Application Failover correctly implemented you get your best uptime numbers. You also get the highest price tag.

    For our HA OLTP system, we used two servers in an Active-Passive HACMP configuration. If server A crashes, server B assume the identitiy of server A and automatically start the Oracle services. To the applications, it looks like server A went away and then came back a few minutes later. There is a 5 - 10 minute outage while the services are restarted and the aplication reconnects. This set-up uses shared disk, which leads to a single point of failure. To minimize this risk, we use EMC disk.

    Standby is another solution. It requires two seperate servers and doubles the amount of disk needed. The caveat is, recovery time is sometimes longer. The application has to enough to go to a different server to find the database.

    If I had to sum this up, I guess I'd say that the closer you have to get to 99.999% uptime, the more expensive the solution. You must start at the business end and see if the cost of an outage justifies the cost of the solution.

    In our case, HACMP was cost justifiable. OPS was not. So, we went with HACMP. I would never choose the solution based on ease of management. I would always work within the guidelines of the business areas needs and provide them the best possible solution based on what they can justify and afford.

    If you send my your e-mail, I can forward you some white-papers on the subject. These white-papers discuss pros and cons of various solutions much better than I can in this posting. Let me know if you are interested.

    HTH,
    Patrick

  5. #5
    Join Date
    Mar 2002
    Location
    Ireland
    Posts
    181
    Hi Patrick,
    The white papers would be useful - breen@o2.ie.
    I am leaning towards clustering software and if not multimaster replication.
    Again - thanks,
    Breen.

  6. #6
    Join Date
    Sep 2002
    Location
    ITALY
    Posts
    53

    Re: 2 box failover / clustering solution for oracle

    Originally posted by Breen
    Hi gentlefolk,
    I was wondering if anybody has any thoughts on the following:
    I will have 2 solaris boxes available to me (E280 or E420) and I want to put in place an easy to use redundency / failover solution for my Oracle Database.
    I have Oracle Enterprise Addition 8i (8.1.7.4) - I am looking at multimaster replication at the moment but was wondering if people have some better ideas.
    Thanks for any ideas or help on the issue,
    Breen.
    State the business needs:
    1. can the business survive a DB out-of-service? for how much time ?
    2. how much data-loss can your business afford ?

    Keep in mind that those questions point out two stategic failure patterns:
    1) zero out-of-service requires RAC
    2) zero data loss needs some kind of replication/standby-database.

    Both solutions can co-exist.

    best regards.
    Franco Ceotto
    SIPTI srl
    OnSite Services
    9i OCP DBA, Performance Engineer

Posting Permissions

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