[OpenIndiana-discuss] sas multipath

Jim Klimov jimklimov at cos.ru
Tue Feb 12 13:08:23 UTC 2013


On 2013-02-12 13:41, Randy S wrote:
>
> Hi,
>
> since we are trying this on 151a7, would there be a way to transfer the mpt driver from a1 to a7, since it seems to work in a1? Or am I thinking too simplistic about this as I am not a programmer...
>

Well, it seems now that you are better experts on this subject than
myself, so I might only offer a few "bystander" advices:

1) What is your pool version? Is it possible to boot it with i.e. an
    oi_151a1 LiveDVD and test whether its driver passes your tests?

2) Alternately, it is likely that from this DVD you can indeed take
    the driver files and try to switch them for the files in the a7
    installation (in an alternate BE and/or renaming originals to
    backup names, of course). I did use this trick for e1000 drivers
    a few years back when something got broken in a newer release;
    the driver ABI should be stable enough in Solaris to allow such
    in-place replacements, and if not - you can roll back.

# find /kernel | grep mpt_sas
/kernel/drv/amd64/mpt_sas
/kernel/drv/mpt_sas.conf
/kernel/drv/mpt_sas
/kernel/kmdb/amd64/mpt_sas
/kernel/kmdb/mpt_sas

3) It is possible that your and Jason's hardware differs on any level
    (including HBA chipset or firmware revision, storage wiring, etc.)
    such that both versions of the driver would work properly for him,
    and fail for you.

4) Finally, if you do find that a1 works for you and a7 doesn't, it
    may be a relatively simple task for a programmer (on bounty or
    otherwise enticed) to compare the changes made between releases
    and find the possible cause of regression - if any. I am not sure
    I am qualified to do this, unless some apparent coding error pops
    out in the diffs - more likely there is some interaction with the
    firmware commands, and one needs to be intimate with them - in
    different revisions of the compatible hardware and firmware - to
    notice the problem causes...

5) It may also be possible that the problem was (or will soon be)
    fixed in later commits to the illumos gate - after the oi_151a7
    LiveDVD and repo were built. In this case it is quite possible
    rebuild the illumos-gate and install it as a new alternate boot
    environment "over" the remainder of the installed OI OS, updating
    the kernel and driver bits and core software included in this gate.

HTH,
//Jim Klimov

> thanks,
>
> Randy
>
>> CC: openindiana-discuss at openindiana.org
>> From: jason at broken.net
>> Date: Wed, 6 Feb 2013 20:29:15 -0800
>> To: openindiana-discuss at openindiana.org
>> Subject: Re: [OpenIndiana-discuss] sas multipath
>>
>>
>> randy,
>>
>> if you do absolutely nothing special - meaning no stmsboot or anything else - the mpt2 driver will default to round robin. failover is sub-sub second.
>>
>> my configuration works like this and is based on 151a1.  if you undo all the old school fancy stuff it will "just work"
>>
>> thanks,
>> j.
>>
>> Sent from Jasons' hand held
>>
>> On Feb 6, 2013, at 4:03 AM, Randy S <sim.ple at live.nl> wrote:
>>
>>>
>>> Hi all,
>>>
>>> I have a test system running with OIa7, very much ram, 2 LSI SAS 9207-4i4e hba's, 1 sm jbod containing seagate 3Tb sas drives.
>>>
>>> I have enabled the multipathing with stmsboot and set it to use logical-blocksize, since round-robin symmetric-option isn't functioning.
>>>
>>> During tests I can see that the system sees the two path's  with mpathadm and simple dd tests show that all disks are all accessed at almost maximum speed with minimum latency according to iostat.
>>> We started to test with 4 sas disks in 2 striped mirrors.
>>>
>>> All good, but shouldn't it be possible to pull one of the sas cables without any disruption to the storage, forcing it to use only one path? (this way simulating a broken hba).
>>> When we do this, the whole system grinds to a screeching halt, being almost unresponsive.
>>>
>>> What am I doing wrong. Maybe somebody can point in in the right direction to troubleshoot.
>>>
>>> Rgrds,
>>>
>>> Randy
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> OpenIndiana-discuss mailing list
>>> OpenIndiana-discuss at openindiana.org
>>> http://openindiana.org/mailman/listinfo/openindiana-discuss
>>
>> _______________________________________________
>> OpenIndiana-discuss mailing list
>> OpenIndiana-discuss at openindiana.org
>> http://openindiana.org/mailman/listinfo/openindiana-discuss
>   		 	   		
> _______________________________________________
> OpenIndiana-discuss mailing list
> OpenIndiana-discuss at openindiana.org
> http://openindiana.org/mailman/listinfo/openindiana-discuss
>


-- 


+============================================================+
|                                                            |
| Климов Евгений,                                 Jim Klimov |
| технический директор                                   CTO |
| ЗАО "ЦОС и ВТ"                                  JSC COS&HT |
|                                                            |
| +7-903-7705859 (cellular)          mailto:jimklimov at cos.ru |
|                        CC:admin at cos.ru,jimklimov at gmail.com |
+============================================================+
| ()  ascii ribbon campaign - against html mail              |
| /\                        - against microsoft attachments  |
+============================================================+






More information about the OpenIndiana-discuss mailing list