Topic
  • 4 replies
  • Latest Post - ‏2012-09-26T11:46:06Z by SteveIves
SteveIves
SteveIves
27 Posts

Pinned topic Unable to run TSA Multiplatforms and OCFS2 on the same node?

‏2012-09-18T09:22:50Z |
We seem to have a fundamental incompatibility between the TSA MP samadapter and OCFS2, due to both of them needing to use the Linux Watchdog timer,

Has anyone else experienced problems in this area?

Thanks, Steve
Updated on 2012-09-26T11:46:06Z at 2012-09-26T11:46:06Z by SteveIves
  • SteveIves
    SteveIves
    27 Posts

    Re: Unable to run TSA Multiplatforms and OCFS2 on the same node?

    ‏2012-09-18T14:26:22Z  
    Apparently the conflict is between the RSCT Watchdog, which is used when the samadapter is automated, to monitor the IBM.ServiceIP resources, and the OCFS2 SBD Watchdog. It would appear that the two cannot co-exist.

    Steve
  • Enrico_Joedecke
    Enrico_Joedecke
    111 Posts

    Re: Unable to run TSA Multiplatforms and OCFS2 on the same node?

    ‏2012-09-18T16:26:09Z  
    Hello Steve,

    the cluster infrastructure (RSCT) uses the watchdog provided with the SuSE Linux, there is no 'own' RSCT watchdog. This watchdog however can only be used by one entity, but it is already used by OCFS2. When RSCT tries to enable it, it detects it already in use. An attempt to un-load the watchdog fails, and so RSCT can't use it, and finally is stuck.
    The watchdog is used for 'critical resource protection'. A critical resource is any resource that must never run on more than one node at any time, e.g. an IP address. The samadapter resource is also defined as 'critical', the information is specified within the resource attribute 'ProtectionMode', which is 1 for critical resources, and 0 for non-critical resources.

    Thank you
  • SteveIves
    SteveIves
    27 Posts

    Re: Unable to run TSA Multiplatforms and OCFS2 on the same node?

    ‏2012-09-18T19:07:24Z  
    Hello Steve,

    the cluster infrastructure (RSCT) uses the watchdog provided with the SuSE Linux, there is no 'own' RSCT watchdog. This watchdog however can only be used by one entity, but it is already used by OCFS2. When RSCT tries to enable it, it detects it already in use. An attempt to un-load the watchdog fails, and so RSCT can't use it, and finally is stuck.
    The watchdog is used for 'critical resource protection'. A critical resource is any resource that must never run on more than one node at any time, e.g. an IP address. The samadapter resource is also defined as 'critical', the information is specified within the resource attribute 'ProtectionMode', which is 1 for critical resources, and 0 for non-critical resources.

    Thank you
    Thanks Enrico,

    So the only options we hare are:

    Disable samadapter automation on these machines, exposing us to a failure of the node the samadapter is running on, whereby we'd have to manually restart the samadapter on another node

    or

    Create a pair of machines for each TSA MP domain on which the samadapter is automated, which don't have any OCFS2 filesystems attached.

    Regards,

    Steve
  • SteveIves
    SteveIves
    27 Posts

    Re: Unable to run TSA Multiplatforms and OCFS2 on the same node?

    ‏2012-09-26T11:46:06Z  
    • SteveIves
    • ‏2012-09-18T19:07:24Z
    Thanks Enrico,

    So the only options we hare are:

    Disable samadapter automation on these machines, exposing us to a failure of the node the samadapter is running on, whereby we'd have to manually restart the samadapter on another node

    or

    Create a pair of machines for each TSA MP domain on which the samadapter is automated, which don't have any OCFS2 filesystems attached.

    Regards,

    Steve
    Problem solved:

    Make the samadapter and samadapter-ip non-critical:

    sudo chrsrc -s "Name like 'samadapter' & ResourceType=1" IBM.Application ProtectionMode=0
    sudo chrsrc -s "Name like 'samadapter-ip' & ResourceType=1" IBM.ServiceIP ProtectionMode=0

    Not ideal, but all that we can apparently do in this situation.

    Steve