Results 1 to 7 of 7
  1. #1
    Join Date
    Jul 2004
    Posts
    306

    Unanswered: HADR Testing - What do you do?

    DB2 8.2 FP9 on AIX 5.3
    Hey all I've got HADR running for several databases using ASYNC. The 2 sites are far apart and connected via a 4Mb link.

    What I'd like to do is get some ideas for some testing of this setup to demonstrate to clients that it is working.
    The two options I can see are:

    1) Break HADR, rollforward the standby systems and give them to the clients to test, then restore all the DB's and setup and sync all the systems again after testing but this means I have no DR during the testing period.

    2) Perform a manual takeover outside business hours and allow the clients to perform non critical operations in the DR systems then takover again at the end of "testing". Obviously the downside to this is that they aren't testing the DB's really just the failover... and they can't do anything that they don't want to end up in their prod systems.

    Any thoughts? What have I missed? What do you guys do?

    Also if anyone know whether a 4Mb link is sufficient for near-sync...monitoring shows me that on ASYNC I'm using on average between 1-3 Mbit with several 4Mbit spikes each day....
    Last edited by meehange; 09-11-07 at 23:30. Reason: change title

  2. #2
    Join Date
    May 2003
    Location
    USA
    Posts
    5,737
    The overall amount of network traffic should be about the same regardless of synch mode, since you are having to send the log data to the standby server in each scenario. The difference is in response time of the commit transaction on the primary server when using synch or near-synch mode.
    M. A. Feldman
    IBM Certified DBA on DB2 for Linux, UNIX, and Windows
    IBM Certified DBA on DB2 for z/OS and OS/390

  3. #3
    Join Date
    Jul 2004
    Posts
    306
    Quote Originally Posted by Marcus_A
    The overall amount of network traffic should be about the same regardless of synch mode, since you are having to send the log data to the standby server in each scenario. The difference is in response time of the commit transaction on the primary server when using synch or near-synch mode.
    Yea I'm just worried that at those times that we hit the 4Mbit mark that commits on primary will be slowed significantly. When we first set it up we were on 2Mbit and we experienced the nasty side effect of "congestion" on the Standby/TCPIP Pipe which let to the logging on primary grinding to a halt, sometimes for several minutes at a time :/
    Increasing the bandwidth and the buffers seems to have fixed that, so I'm looking at moving up to near-sync now

  4. #4
    Join Date
    Aug 2001
    Location
    UK
    Posts
    4,650
    To confirm, HADR is working, you have to demostrate two things :

    a) When takeover happens, all transactions committed on the primary are available in the secondary.

    b) When you switch back to the primary, all transactions that happened on the secondary are applied to primary

    Therefore, I would go with option 2.

    Sathyaram

    Quote Originally Posted by meehange
    DB2 8.2 FP9 on AIX 5.3
    Hey all I've got HADR running for several databases using ASYNC. The 2 sites are far apart and connected via a 4Mb link.

    What I'd like to do is get some ideas for some testing of this setup to demonstrate to clients that it is working.
    The two options I can see are:

    1) Break HADR, rollforward the standby systems and give them to the clients to test, then restore all the DB's and setup and sync all the systems again after testing but this means I have no DR during the testing period.

    2) Perform a manual takeover outside business hours and allow the clients to perform non critical operations in the DR systems then takover again at the end of "testing". Obviously the downside to this is that they aren't testing the DB's really just the failover... and they can't do anything that they don't want to end up in their prod systems.

    Any thoughts? What have I missed? What do you guys do?

    Also if anyone know whether a 4Mb link is sufficient for near-sync...monitoring shows me that on ASYNC I'm using on average between 1-3 Mbit with several 4Mbit spikes each day....
    Visit the new-look IDUG Website , register to gain access to the excellent content.

  5. #5
    Join Date
    Jul 2004
    Posts
    306
    Quote Originally Posted by sathyaram_s
    To confirm, HADR is working, you have to demostrate two things :

    a) When takeover happens, all transactions committed on the primary are available in the secondary.

    b) When you switch back to the primary, all transactions that happened on the secondary are applied to primary

    Therefore, I would go with option 2.

    Sathyaram
    I agree, but the problem there is selling the idea to management and clients. There is no element of trust in HADR as it is a new concept to them. I can see that the logfiles are being applied and committed via the monitoring but in order to demonstrate that via the front end you have to essentially use the production systems in the test. Which is why I'm wondering what everyone else does?

  6. #6
    Join Date
    Jul 2004
    Posts
    306
    Anyone got anything more to say on this one?

  7. #7
    Join Date
    Jun 2007
    Posts
    12
    On a recent HADR demo I did, we added new data to the primary, did a manual takeover and then selected the inserted data from the new primary. After a takeover to return to the original primary, we deleted the new test data. The users were very impressed by how transparent the whole process was to the end users.

Posting Permissions

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