CentOS - Mar 2016 - Odd behavior of a CentOS 7 box after repair of an external RAID

Valeri Galtsev wrote:
>
> On Fri, March 25, 2016 9:55 am, m.roth at 5-cent.us wrote:
>> I don't think I've seen this with CentOS 6 or 5, but I had to
repair an
>> external RAID box this morning. The server, running CentOS 7, has an
>> LSI HBA card in it, and it's presented to the system as /dev/sdb.
>> I shut off the RAID controller, powered it off,
>
> You did unmount filesystem on the machine before that, right?
Of course!
>
>> did the repair, then plugged everything back in (that includes the two
>> fiber cables and the ethernet), and brought the RAID back up. I even
>> used its web page to restart the controller (in the RAID box.
>>
>> I couldn't remount the drive until I rebooted the server.
>
> Before powering off/disconnecting RAID box, did you remove resembling SCSI
> device? this is done if my memory doesn't fail me by doing this:
>
> echo 1 > /sys/class/scsi_device/6:0:0:0/device/delete
No - I've never done that. With hot-swap drive bays, the system sees it
going out, and says so, and then the drive goes away, or becomes
inactive... but when I shove one back in, it sees it, though it may make
it as /dev/sd[letter+1] (that is, if it had been /dev/sdb, it's now
/dev/sdc).
>
> (replace "6:0:0:0" with your device's address). If you
didn't remove the
> device before powering it off, I'm not surprised. If you did remove the
> device before powering it off, I'm puzzled ;-)
As I said, I expected the scsi-rescan-bus to clear it up, but though it
saw it, nothing changed.
<snip>
     mark

On Fri, March 25, 2016 10:45 am, m.roth at 5-cent.us
wrote:
> Valeri Galtsev wrote:
>>
>> On Fri, March 25, 2016 9:55 am, m.roth at 5-cent.us wrote:
>>> I don't think I've seen this with CentOS 6 or 5, but I had
to repair an
>>> external RAID box this morning. The server, running CentOS 7, has
an
>>> LSI HBA card in it, and it's presented to the system as
/dev/sdb.
>>> I shut off the RAID controller, powered it off,
>>
>> You did unmount filesystem on the machine before that, right?
>
> Of course!
>>
>>> did the repair, then plugged everything back in (that includes the
two
>>> fiber cables and the ethernet), and brought the RAID back up. I
even
>>> used its web page to restart the controller (in the RAID box.
>>>
>>> I couldn't remount the drive until I rebooted the server.
>>
>> Before powering off/disconnecting RAID box, did you remove resembling
>> SCSI
>> device? this is done if my memory doesn't fail me by doing this:
>>
>> echo 1 > /sys/class/scsi_device/6:0:0:0/device/delete
>
> No - I've never done that. With hot-swap drive bays, the system sees it
> going out, and says so, and then the drive goes away, or becomes
> inactive... but when I shove one back in, it sees it, though it may make
> it as /dev/sd[letter+1] (that is, if it had been /dev/sdb, it's now
> /dev/sdc).
Then we will need real expert to chime in on this...
>>
>> (replace "6:0:0:0" with your device's address). If you
didn't remove the
>> device before powering it off, I'm not surprised. If you did remove
the
>> device before powering it off, I'm puzzled ;-)
>
> As I said, I expected the scsi-rescan-bus to clear it up, but though it
> saw it, nothing changed.
Did you do scsi-rescan-bus after you disconnected the device, and then
again after you connected it back? I'm just shooting in a dark, but the
first one probably will remove the device that disappeared, and the second
will add the device back. And it may do nothing about devices that exist
on the bus, even though they disappeared temporarily, then re-appeared. I,
simple person, knowing not all devices/scsi controllers support hot swap,
would just remove the device as I said in previous post before
disconnecting (an rescan scsi bus after connecting it back).

Valeri
> <snip>
>      mark
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> https://lists.centos.org/mailman/listinfo/centos
>

++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev
Sr System Administrator
Department of Astronomy and Astrophysics
Kavli Institute for Cosmological Physics
University of Chicago
Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++

Valeri Galtsev wrote:
> On Fri, March 25, 2016 10:45 am, m.roth at 5-cent.us wrote:
>> Valeri Galtsev wrote:
>>> On Fri, March 25, 2016 9:55 am, m.roth at 5-cent.us wrote:
>>>> I don't think I've seen this with CentOS 6 or 5, but I
had to repair
>>>> an external RAID box this morning. The server, running CentOS
7, has an
>>>> LSI HBA card in it, and it's presented to the system as
/dev/sdb.
>>>> I shut off the RAID controller, powered it off,
<snip>
>> No - I've never done that. With hot-swap drive bays, the system
sees it
>> going out, and says so, and then the drive goes away, or becomes
>> inactive... but when I shove one back in, it sees it, though it may
make
>> it as /dev/sd[letter+1] (that is, if it had been /dev/sdb, it's now
>> /dev/sdc).
>
> Then we will need real expert to chime in on this...
Yeah, I was wondering about, what, dbus? dracut?
<snip>
>> As I said, I expected the scsi-rescan-bus to clear it up, but though it
>> saw it, nothing changed.
>
> Did you do scsi-rescan-bus after you disconnected the device, and then
> again after you connected it back? I'm just shooting in a dark, but the
No. I didn't think of doing that, because I thought that once I'd shut
the
RAID controller down, and powered it off, the system would have noticed.
> first one probably will remove the device that disappeared, and the second
> will add the device back. And it may do nothing about devices that exist
That might have worked; certainly, it's a thought to keep in mind for the
future (he says, pushing it onto the stack).
<snip>
       mark