1. No Profile Picture
    Registered User
    Devshed Newbie (0 - 499 posts)

    Join Date
    Jun 2011
    Rep Power

    Corrupt FB 2.5 database


    We have an issue where a database hangs the server when
    > a backup is performed (Hangs on a specific table)
    > selecting * or count(1) from a specific table
    > or viewing data that is related to the table (FKs, etc)

    We could browse the table to a certain point (using IBExpert) however after about 2900 records the machine just spikes and hangs.
    Performing a gfix -m does not work, and the validation reports back Record level errors = 4. (No matter how many times we run gfix -m, -v, etc.

    The Firebird.log file reports back these types of messages(Multiple instances of each of the below messages):
    -Relation has 91631 orphan backversions (9214273 in use) in table BINS (137) - {Which is apparently just a warning}
    -Unable to complete network request to host "MHPLZA1".
    Error reading data from the connection.
    -INET/inet_error: read errno = 10054
    -SERVER/process_packet: broken port, server exiting
    -Shutting down the server with 1 active connection(s) to 1 database(s), 0 active service(s) - {If we leave the backup to run while hanging, it eventually logs this error message}

    The table is question has about 7000 records only
    The Firebird version is 2.5 Classic Server x64 install
    The OS is Windows Server 2008
    This is a virtual machine (VMWare) running on a massive server. (Does anyone have issues with VMs and Firebird?)
    We have the same setup running fine on other servers (However they are not virtual machines)

    Is there anyway to pin point the issue and or the cause?

    Thanks in advance for any and all help!

  2. #2
  3. No Profile Picture
    Contributing User
    Devshed Newbie (0 - 499 posts)

    Join Date
    Nov 2009
    Rep Power
    Interesting post!

    We are running:
    Firebird 2.04 superserver on a XEN visualization --> Debian 5.

    Since a software update we are also encountering database hangs.
    Some say: it could be XEN, some say: it could be FB super, some say it is the application, some say bull****.

    Cause is still unknown. Errors are different. Prob. multiple problems...

    One error is saying that a deadlock occurs...

    So about deadlocks. Is this a database issue or an application issue?
    I say application. The application should perform deadlock prevention or resolution right?

    Currently I am developing some test cases. New hardware, stress scripts, different VM's/ OS, newest FB version super/classic, etc, etc.
    To make hard conclusions you can only change on thing at the time and it is very difficult to make your test situation the same as your production server.

    In case the conclusion will be something like: Don't run FB on a VM, I will inform you.

    Some tips:

    - Try this: http://www.firebirdnews.org/?p=1881
    - Try Linux
    - Try another VM like KVM.
    - Try MySQL or PostGreSQL if possible

    I can suggest and provide more details, but I am not allowed to do.
    This is a kind of complex problem with too many variables.
    That's the reason I want to test with different scenario's. A very time consuming and difficult task. But it is needed. I am tired of all the different suggestions, lack of documentation, lack of answers, lack of support, lack of log messages.

    Lets test it and solve it our selfs!!!
    And lets keep in touch.

    Another tip:

    Last edited by rapgame; August 8th, 2011 at 01:59 PM.

IMN logo majestic logo threadwatch logo seochat tools logo