Re: SSA Loop Verification Question (graphics)

From: Bill Verzal (BVerzal_at_KOMATSUNA.COM)
Date: 04/25/03

  • Next message: Erin Michaud: "NFS help"
    Date:         Fri, 25 Apr 2003 09:57:24 -0500
    To: aix-l@Princeton.EDU
    
    

    Thanks for your thoughts.

    Indeed, all 5 nodes would need access to all disks at one time or another,
    so it needs to stay in that mode. In terms of outages, this is just a
    backup pool, so if there are hosts down, it is of no concern, unless if
    course backups are running. We've never had an instance where there were 2
    nodes down, but it could happen of course.

    Thanks again,

    Bill.
    ___________________________________________
    Bill Verzal
    AIX Admin/Komatsu America
    (847) 970-3726 (v)

    |---------+---------------------------->
    | | "Green, Simon" |
    | | <Simon.Green@EU.A|
    | | LTRIA.COM> |
    | | Sent by: IBM AIX |
    | | Discussion List |
    | | <aix-l@Princeton.|
    | | EDU> |
    | | |
    | | |
    | | 04/25/2003 09:33 |
    | | AM |
    | | Please respond to|
    | | IBM AIX |
    | | Discussion List |
    | | |
    |---------+---------------------------->
    >-------------------------------------------------------------------------------------------------------------------------------|
      | |
      | To: aix-l@Princeton.EDU |
      | cc: |
      | Subject: Re: SSA Loop Verification Question (graphics) |
    >-------------------------------------------------------------------------------------------------------------------------------|

    I would switch the cables round on SAP80e, (swap A1 with A2; B1 with B2),
    so
    that throughout the loop, an A1 connects to an A2 etc. It'll work fine as
    it is, but if you run a maymap or similar from SAP80e it will show the
    disks
    in the opposite order to one run on any other node. It's no extra effort
    for you and may save future confusion.

    Give some though to the bypass cards. I think you need to make sure that
    the 1/16 and 8/9 cards are set to forced in-line, given that you've got so
    many nodes in the loop.

    Be aware that if you have two nodes down there is a risk of some disks
    being
    isolated. Unless you have a compelling reason to have one big, shared loop
    I would strongly recommend splitting it up so that you have no more than
    two
    nodes in any one loop. Split drawers up if you need to, but that'll be a
    lot less trouble than having so many nodes in the loop, in my experience.

    Print some cable plans off, and trace the loops with various combinations
    of
    nodes down and bypass configurations. You might consider failures of SSA
    drawers as well. I usually put the cable plan in a clear plastic sleeve
    and
    trace over it with dry-wipe pens in various colours.

    If you need the single loop, an alternative configuration, (A loop only,
    for
    the front half of the drawers), would be to connect each node to ports 4
    and
    5, then run cables between the drawers from 8 to 1; similar for the B loop.
    That way, if a node drops out the whole loop will remain intact, thanks to
    the bypass cards. (Still need to keep the back and front discrete.) It
    will even permit access to all disks when two or more nodes are down. The
    downside is that it needs more cables.

    Where possible, keep the disks for a particular node in the drawer to which
    it is directly connected. That will help performance, and also make it
    easier if you decide to split the loop later on. You probably don't have a
    lot of choice with that at the moment, though.

    Simon Green
    Altria ITSC Europe s.a.r.l.

    AIX-L Archive at http://marc.theaimsgroup.com/?l=aix-l&r=1&w=2
    AIX FAQ at http://www.faqs.org/faqs/aix-faq/

    N.B. Unsolicited email from vendors will not be appreciated.

    > -----Original Message-----
    > From: Bill Verzal [mailto:BVerzal@KOMATSUNA.COM]
    > Sent: 25 April 2003 14:41
    > To: aix-l@Princeton.EDU
    > Subject: SSA Loop Verification Question (graphics)
    >
    >
    > All,
    >
    > I have attached a JPG image of a proposed SSA loop
    > configuration. I would
    > like your feedback on it if you don't mind. Here is the scenario...
    >
    > I have (5) 7133-020 drawers and (5) hosts. There are 80 disks in the
    > drawers (all 5 drawers are full).
    >
    > Currently, I have 36 devices in loop A and 52 in loop B (this
    > is just the
    > way it is now - I did not wire it up).
    > Loop A is broken down 32 drives and 4 hosts (5 counting the
    > host the map
    > was drawn from).
    > Loop B is broken down 48 drives and 4 hosts (5 counting the
    > host the map
    > was drawn from).
    > After adding the last 8 drives I am getting
    > SSA_DEGRADED_ERROR on my hosts.
    >
    > Please take a look at the JPG and offer your opinions.
    >
    > The goal of the proposed changes is to keep the front 8
    > drives isolated
    > from the read 8, as they contain different data. Currently,
    > front and rear
    > drives cross over all over the place.


  • Next message: Erin Michaud: "NFS help"

    Relevant Pages

    • Sunfire V480 ,Solaris 9 VxVm 3.5 and X6727
      ... The X6727 is used for the double loop to the internal disks. ... The installation went in the following order. ... to try to encapsulate the bootdisk at reboot time it ...
      (SunManagers)
    • SSA Cabling and Adapter problem
      ... A2 -> Port 16 ... So the A loop is fine, ... The server is an S7A with two I/O drawers, ... There is a seventh adapter in the server, not connected to any disks at all, ...
      (AIX-L)
    • Re: hdisks missing
      ... So some disks are used by one node, ... After cfgmgr the hdisks are still ... What sort of SSA drawer is it? ... deliberate changes to the SSA loop it's possible that the cables were ...
      (AIX-L)
    • Re: SSA Loop Verification Question (graphics)
      ... many nodes in the loop. ... Be aware that if you have two nodes down there is a risk of some disks being ... Print some cable plans off, and trace the loops with various combinations of ... drawers as well. ...
      (AIX-L)
    • Re: [PATCH] iSeries virtual disk
      ... > easily happen on the i used as loop counter. ... In theory you can hot add virtual disks to an iSeries partition (up to 64 ... viodasd_max_disk will never decrease, because you cannot remove a virtual ...
      (Linux-Kernel)