SUMMARY: T3 disk array LUNs & redundancy

I received several replies confirming that a T3 partner pair is
indeed fully redundant, even if one of the two controller boards
should fail.  The remaining T3's controller can still provide
full access to all four LUNs.

Detailed responses follow...

Doug Otto came straight to the point:

    Short answer - yes.
 
    -----Original Message-----
    ~snip~   
 
    ...on the other hand, Sun also says that a single
    controller can only support two LUNs total, so if one
    fails, can the other controller still give you access
    to all four LUNs?
    ~snip    


Jim Ennis related this experience:

    We had a controller fail in a partner pair (master).  The
    slave became the master and allowed full access to all four
    LUNS (two per T300) that we had defined.  There was a
    momentary loss of connectivity during the switchover but
    our Oracle DB kept on running just fine through the failover.


Sampath Kari provided this detailed explanation:

    In case of one controller failure, all 4 LUNS are
    accessible via the good controller on other T3 in the 
    partner group configuration.

    Look at format, you find 4 luns from each controller.
    (if you have any path failover software such as
    veritas volume manager dmp).

    If you have the partner group at your place, you can
    test it by just pulling a cable to one controller and
    verifying access to all 4 luns from host via one 
    controller itself.

    Note that cache will be in write through mode during
    the time one of the controller is bad. 

+--- original question ---
| 
| To: sunmanagers@sunmanagers.org 
| Subject: T3 disk array LUNs & redundancy
|  
| This one doesn't have a definitive answer in any of Sun's online
| specifications, FAQs, configuration guides or service manuals; I've
| looked for hours without success.   
|  
| Say you've got two T3 disk arrays interconnected as a partner pair.
|  
| A partner pair consists of two T3 disk trays linked together through
| their unit interconnects.  Each tray has its own controller and FC
| interface.  Supposedly, you have a great deal of redundancy in the
| partner pair setup; if any one component fails, there's an alternate  
| path or a backup component.  but...
|  
| If you have two LUNs configured in each tray - a total of four LUNs,
| and the controller in one tray fails, what happens? 
|  
| On one hand, Sun says the controller in either tray can provide access
| to all the disks in both trays; on the other hand, Sun also says that
| a single controller can only support two LUNs total, so if one fails,
| can the other controller still give you access to all four LUNs?
|  
|  
|  
|  
| Here are direct quotes from Sun materials...
|  
| 1. "Should a controller fail, the alternate controller provides
| access to all 18 drives."
|  
| 2. "Controller Card
|     One RAID controller card/tray   RAID level 0, 1(1+0), 5
|                                     1 or 2 LUNs per enclosure, total 4
|                                     LUNs per unit"
|  
| 3. "DESCRIPTION                 StorEdge A5200    StorEdge T3 Array"
|       [...]
|    "Max. #LUNs per disk tray    22                2 per Controller Unit"
|  
|  
| [1] http://www.sun.com/storage/t3es/t3es_datasheet.pdf, p.1, 4th sentence
|  
| [2] same document, p.2, left column under 'Controller Card' heading.
|  
| [3] http://www.sun.com/storage/t3es/faq.html, answer to question 8.
|  
|  
| Does anyone know what happens in the failure scenario I described?
|  
| Many thanks for your help!
|  
| Regards,
| Michael Maciolek
| 
+--- end of original question ---
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers