Topic
  • 10 replies
  • Latest Post - ‏2014-10-01T20:21:40Z by Shark800
ajaynp
ajaynp
8 Posts

Pinned topic IBM V7000 Cluster

‏2013-11-11T13:50:39Z |

Hi

We want to create cluster between two controller enclosures  Following  are the details of controllers enclosures , Expansion enclosures & Hard drives.

Existing V7000 Storage Details
Controller Enclosure P/N: 2076-324  
1 Expansion enclosure P/N: 2076-212
Hard disks: 600 GB  SAS  23 no's
3TB NL SAS. 11 Nos.
 
New V7000 Storage Details
Controller Enclosure P/N 2076-312
2 Expansion enclosure P/N 2076-212
Hard disks: 4TB NL SAS. 24 No's
 
Also there are some more queries.
 
1)   As per IBM information Centre New Controller enclosure should not initialize for adding it to the cluster. If the controller enclosure firmware is different how it can be added to the cluster?

2) Which are the FC ports to be zoned for clustering?

3) How we can update the firmware without initializing the storage?

4)   If we can have some procedures on how new enclosures can be added to the clustered system? We tried Redbooks, which doesn't have detailed information.

5) In this scenario Hard disks capacity of new storage and existing storage is different

Please help

 

Updated on 2013-11-11T13:56:55Z at 2013-11-11T13:56:55Z by ajaynp
  • chriscanto
    chriscanto
    292 Posts
    ACCEPTED ANSWER

    Re: IBM V7000 Cluster

    ‏2013-11-11T14:49:22Z  

    This is the V7000 Information Center page for "Adding another control enclosure into an existing system":

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Ftbrd_qicfgsys_addctlencl_b433xw1.html

    If you follow those steps then the second control enclosure will be added to your existing system.  The software level on the new control enclosure is automatically upgraded/downgraded as required to match the existing system, no action is required on your part for that.

    You should zone at least two FC ports on each node canister for clustering.  It doesn't matter which ones you use.   If you have enough switch ports then you can use all four to maximise bandwidth and redundancy.  The important thing is to make sure there is equal connectivity to each node canister.

  • chriscanto
    chriscanto
    292 Posts
    ACCEPTED ANSWER

    Re: IBM V7000 Cluster

    ‏2013-11-11T15:45:29Z  
    • ajaynp
    • ‏2013-11-11T15:19:20Z

    hi chirs

    thanks for the answers.

    . I have few more doubts.

    1) In the case i have different type of controller enclosure  P/N: 2076-324 &P/N 2076-312 . is this ok to create cluster with these controllers?

    2) Hard disk types are different . 4 TB ,3TB & 600 GB . I  think we cannot create I/O Groups across the enclosures?

    3) in this case  do I Need to Zone the host from both controller enclosures?

    4) if yes each hosts should have 8 paths 

    could you please clarify?

     

    1) It is supported to create a clustered system with different models of control and expansion enclosures, no problems there.

    2) An I/O group is specifically one control enclosure, containing two node canisters.  So when you add the second control enclosure you are adding a second I/O group.   The fact that I/O group may have SAS attachment to different types of drives is no problem.

    3) You only need to zone a host to node ports on the control enclosure(s) it will access volumes from.

    4) Zoning guidelines can be found in the Information Center here:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Fsvc_zoningguidelines_22n6zd.html

  • chriscanto
    chriscanto
    292 Posts

    Re: IBM V7000 Cluster

    ‏2013-11-11T14:49:22Z  

    This is the V7000 Information Center page for "Adding another control enclosure into an existing system":

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Ftbrd_qicfgsys_addctlencl_b433xw1.html

    If you follow those steps then the second control enclosure will be added to your existing system.  The software level on the new control enclosure is automatically upgraded/downgraded as required to match the existing system, no action is required on your part for that.

    You should zone at least two FC ports on each node canister for clustering.  It doesn't matter which ones you use.   If you have enough switch ports then you can use all four to maximise bandwidth and redundancy.  The important thing is to make sure there is equal connectivity to each node canister.

  • ajaynp
    ajaynp
    8 Posts

    Re: IBM V7000 Cluster

    ‏2013-11-11T15:19:20Z  

    This is the V7000 Information Center page for "Adding another control enclosure into an existing system":

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Ftbrd_qicfgsys_addctlencl_b433xw1.html

    If you follow those steps then the second control enclosure will be added to your existing system.  The software level on the new control enclosure is automatically upgraded/downgraded as required to match the existing system, no action is required on your part for that.

    You should zone at least two FC ports on each node canister for clustering.  It doesn't matter which ones you use.   If you have enough switch ports then you can use all four to maximise bandwidth and redundancy.  The important thing is to make sure there is equal connectivity to each node canister.

    hi chirs

    thanks for the answers.

    . I have few more doubts.

    1) In the case i have different type of controller enclosure  P/N: 2076-324 &P/N 2076-312 . is this ok to create cluster with these controllers?

    2) Hard disk types are different . 4 TB ,3TB & 600 GB . I  think we cannot create I/O Groups across the enclosures?

    3) in this case  do I Need to Zone the host from both controller enclosures?

    4) if yes each hosts should have 8 paths 

    could you please clarify?

     

  • chriscanto
    chriscanto
    292 Posts

    Re: IBM V7000 Cluster

    ‏2013-11-11T15:45:29Z  
    • ajaynp
    • ‏2013-11-11T15:19:20Z

    hi chirs

    thanks for the answers.

    . I have few more doubts.

    1) In the case i have different type of controller enclosure  P/N: 2076-324 &P/N 2076-312 . is this ok to create cluster with these controllers?

    2) Hard disk types are different . 4 TB ,3TB & 600 GB . I  think we cannot create I/O Groups across the enclosures?

    3) in this case  do I Need to Zone the host from both controller enclosures?

    4) if yes each hosts should have 8 paths 

    could you please clarify?

     

    1) It is supported to create a clustered system with different models of control and expansion enclosures, no problems there.

    2) An I/O group is specifically one control enclosure, containing two node canisters.  So when you add the second control enclosure you are adding a second I/O group.   The fact that I/O group may have SAS attachment to different types of drives is no problem.

    3) You only need to zone a host to node ports on the control enclosure(s) it will access volumes from.

    4) Zoning guidelines can be found in the Information Center here:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Fsvc_zoningguidelines_22n6zd.html

  • ajaynp
    ajaynp
    8 Posts

    Re: IBM V7000 Cluster

    ‏2013-11-12T05:30:26Z  

    1) It is supported to create a clustered system with different models of control and expansion enclosures, no problems there.

    2) An I/O group is specifically one control enclosure, containing two node canisters.  So when you add the second control enclosure you are adding a second I/O group.   The fact that I/O group may have SAS attachment to different types of drives is no problem.

    3) You only need to zone a host to node ports on the control enclosure(s) it will access volumes from.

    4) Zoning guidelines can be found in the Information Center here:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Fsvc_zoningguidelines_22n6zd.html

    Hi chris,

     

    thank you very much for the info

  • charles1
    charles1
    2 Posts

    Re: IBM V7000 Cluster

    ‏2014-01-09T09:26:14Z  

    1) It is supported to create a clustered system with different models of control and expansion enclosures, no problems there.

    2) An I/O group is specifically one control enclosure, containing two node canisters.  So when you add the second control enclosure you are adding a second I/O group.   The fact that I/O group may have SAS attachment to different types of drives is no problem.

    3) You only need to zone a host to node ports on the control enclosure(s) it will access volumes from.

    4) Zoning guidelines can be found in the Information Center here:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.710.doc%2Fsvc_zoningguidelines_22n6zd.html

    hi Chriscanto,

     

    I have to do the same, and in order to prepare my new installation, can you let me know how many IP address I need?

     

    In fact, I have to install 2 V7000 (2076-124) in cluster mode (2 IO groups).  I assume I need only one management IP address and 4 service IP addresses.

    Can you confirm that point?

    In advance, I thank you very much.

    Cheers,

    Charles

  • chriscanto
    chriscanto
    292 Posts

    Re: IBM V7000 Cluster

    ‏2014-01-09T09:33:17Z  
    • charles1
    • ‏2014-01-09T09:26:14Z

    hi Chriscanto,

     

    I have to do the same, and in order to prepare my new installation, can you let me know how many IP address I need?

     

    In fact, I have to install 2 V7000 (2076-124) in cluster mode (2 IO groups).  I assume I need only one management IP address and 4 service IP addresses.

    Can you confirm that point?

    In advance, I thank you very much.

    Cheers,

    Charles

    You're correct.

    A minimum of one management addresses must be allocated for the system. The first management address is assigned for use on Ethernet port 1. The system assigns the management addresses to one node in the system at any time, to provide access to system configuration and administration functions such as the management CLI and GUI.

    A service address should also be allocated to each node (canister) in the system. The service address of a node is allocated to Ethernet port 1 of the node. The node retains the service IP address at all times, allowing convenient access to node configuration and service functions, such as the service assistant GUI and CLI for that node.

    You might find the following page in the Information Center useful to help with your planning:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.doc%2Ftbrd_completeconfdata.html

  • charles1
    charles1
    2 Posts

    Re: IBM V7000 Cluster

    ‏2014-01-09T14:32:42Z  

    You're correct.

    A minimum of one management addresses must be allocated for the system. The first management address is assigned for use on Ethernet port 1. The system assigns the management addresses to one node in the system at any time, to provide access to system configuration and administration functions such as the management CLI and GUI.

    A service address should also be allocated to each node (canister) in the system. The service address of a node is allocated to Ethernet port 1 of the node. The node retains the service IP address at all times, allowing convenient access to node configuration and service functions, such as the service assistant GUI and CLI for that node.

    You might find the following page in the Information Center useful to help with your planning:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.doc%2Ftbrd_completeconfdata.html

    Hi Chris,

     

    Many thanks for your quick and detailed reply.

    Now I can prepare everything properly.

    Charles

  • Shark800
    Shark800
    3 Posts

    Re: IBM V7000 Cluster

    ‏2014-10-01T19:24:57Z  

    You're correct.

    A minimum of one management addresses must be allocated for the system. The first management address is assigned for use on Ethernet port 1. The system assigns the management addresses to one node in the system at any time, to provide access to system configuration and administration functions such as the management CLI and GUI.

    A service address should also be allocated to each node (canister) in the system. The service address of a node is allocated to Ethernet port 1 of the node. The node retains the service IP address at all times, allowing convenient access to node configuration and service functions, such as the service assistant GUI and CLI for that node.

    You might find the following page in the Information Center useful to help with your planning:

    http://pic.dhe.ibm.com/infocenter/storwize/ic/index.jsp?topic=%2Fcom.ibm.storwize.v7000.doc%2Ftbrd_completeconfdata.html

    Chris,

    I have a client w/ two exsisting production v7000's non-clustered running 7.1.0.6 code.  They are adding SSD enclosures 2 per environment to take advantage of EasyTier, but also want to cluster the systems.  I've experianced new cluster implementations, but not finding any information regarding clustering exsisting systems, or impact of production pools\mdisk that are exsisting if now clustering? 

    Any help or direction is greatly appreicated.

    Respectfully,

    Patrick

     

  • chriscanto
    chriscanto
    292 Posts

    Re: IBM V7000 Cluster

    ‏2014-10-01T20:05:07Z  
    • Shark800
    • ‏2014-10-01T19:24:57Z

    Chris,

    I have a client w/ two exsisting production v7000's non-clustered running 7.1.0.6 code.  They are adding SSD enclosures 2 per environment to take advantage of EasyTier, but also want to cluster the systems.  I've experianced new cluster implementations, but not finding any information regarding clustering exsisting systems, or impact of production pools\mdisk that are exsisting if now clustering? 

    Any help or direction is greatly appreicated.

    Respectfully,

    Patrick

     

    Hello Patrick,

    It's not possible to cluster two existing systems together, i.e. merging two separate systems into one clustered system.

    SVC/V7000 scale-out works by adding 'candidate' nodes/enclosures to the existing system.

    Hope that helps,

    Chris

  • Shark800
    Shark800
    3 Posts

    Re: IBM V7000 Cluster

    ‏2014-10-01T20:21:40Z  

    Hello Patrick,

    It's not possible to cluster two existing systems together, i.e. merging two separate systems into one clustered system.

    SVC/V7000 scale-out works by adding 'candidate' nodes/enclosures to the existing system.

    Hope that helps,

    Chris

    Chris,

    Yes...what I had figured, but had never experianced particular request..  Appreciate the help, and quick response!