For the regular power on detected, are you referring to cold/warm
booting the system and verifying initialization or only power cycling
the 4726?

Here are my test results, TestingRoutine.txt is the procedure I followed
and test results obtained. If I missed something, please let me know.

>From what I could find, the problem with ata1.03 not being detected is
consistent with the port. In my testing, I moved the last drive to the
5th port on the 4726 and the problem carried through. The issue seemed
to only occur when the 4726 was powered on/off, not when it was
unplugged and re-plugged with the power on.

If I have some time on the weekend, I will do a test routine of several
hot un-plugs / re-plugs of the 4726 and see what happens.

Thanks


Tejun Heo wrote:
> Hello,
>
> Andrew Ryder wrote:
>   
>> This might be of interest.. I had turned the drive off for about 20
>> minutes, then turned it back on. ata1.03 is now partially there, it is
>> recognized as a device, but of 0MB in size.
>>
>> All of this is done with the 4726 turned on..
>>
>> Bootlog11.txt -> I unplugged the 4726, waited 10 seconds, and plugged it
>> into the second port on the 3124. All 4 drives are there.
>>
>> Bootlog12.txt -> I unplugged the 4726 again, waited 10 seconds, plugged
>> it back into the first (original) port, and ata1.03 is back..
>>
>> Bootlog13.txt -> I powered the 4726 off, waited 10 seconds, turned it
>> back on. The same drive is missing now.
>>
>> Bootlog14.txt -> I unplugged, waited 10 seconds, and re-plugged the 4726
>> into the same port (first eSATA port) and now ata1.03 is recognized and
>> accessible.
>>
>> I'll make sure to turn on the timestamping for you tomorrow. Would want
>> me to redo this testing so you can have the timestamps also?
>>     
>
> Please swap the third and fourth drive and see whether the problem moves
> with the drive or stays with the port.  Also, please verify regular
> powered on detection (array up and running before detection begins)
> always works.  4726 is very quirky (it ate up a LOT of development /
> testing time) and I guess we'll have to accept some brokenness.
>
> Thanks.
>
>