Am 20.02.2013 01:31, schrieb Phil Turmel:

You forgot to include linux-raid again.  I'm adding them back to the
CC:.  Please always use "reply to all" in your email client.

Sorry.

I will look for your detailed reply tomorrow.

Phil

On 02/19/2013 05:23 PM, Stone wrote:

quoted

Am 19.02.2013 23:08, schrieb Phil Turmel:

quoted

On 02/19/2013 04:31 PM, Stone wrote:

[trim /]

quoted

quoted

[trim /]

ok. my system is a ubuntu 12.04
i can install a older mdadm or a install a old ubuntu like 11.04. there
is a older mdadm on board.

Using the older ubuntu as a LiveCD should be fine--you don't have to
uninistall your current system.

[trim /]

quoted

ok. here my next steps
i find a older mdadm or i install a older ubunt with an older mdadm on
board.
then i stop my md2 device and recreate it with: mdadm --create /dev/md2
--assume-clean --verbose --level=5 --raid-devices=4 /dev/sdc1 /dev/sdd1
missing /dev/sdf1

Yes.  But read all the way through first....

quoted

with a little bit of hope i can open the device.

But *don't* mount it!  Use "fsck -n" after you open it to verify it is
Ok.  If you mount it, and the chunk size is wrong, it will damage your
encrypted filesystem.

quoted

if not. i stop the md2 and recreate it with? with the parameter chunk?
and with what value? do you have a range for me?

The current default is 512.  The old default was 64.  I'd try that if
512 doesn't work.  After that you'll have to guess.

Ok i will test this tomorrow.

quoted

quoted

here the timeout infos:
for x in /sys/block/sd*/device/timeout ; do echo $x ; cat $x ; done
/sys/block/sda/device/timeout
30
/sys/block/sdb/device/timeout
30
/sys/block/sdc/device/timeout
30
/sys/block/sdd/device/timeout
30
/sys/block/sde/device/timeout
30
/sys/block/sdf/device/timeout
30

Ok, these are all Linux default.  30 seconds.

quoted

here the smart infos:

Uh oh.  Two serious issues:

quoted

smartctl -x /dev/sdc1
smartctl 5.41 2011-06-09 r3365 [x86_64-linux-3.2.0-23-generic] (local
build)
Copyright (C) 2002-11 by Bruce Allen,
http://smartmontools.sourceforge.net

[trim /]

quoted

    5 Reallocated_Sector_Ct   PO--CK   200   200   140    -    0
    7 Seek_Error_Rate         -OSR-K   200   200   000    -    0
    9 Power_On_Hours          -O--CK   078   078   000    -    16219
   10 Spin_Retry_Count        -O--CK   100   100   000    -    0
   11 Calibration_Retry_Count -O--CK   100   253   000    -    0
   12 Power_Cycle_Count       -O--CK   100   100   000    -    84
192 Power-Off_Retract_Count -O--CK   200   200   000    -    82
193 Load_Cycle_Count        -O--CK   169   169   000    -    94419
194 Temperature_Celsius     -O---K   114   106   000    -    36
196 Reallocated_Event_Count -O--CK   200   200   000    -    0
197 Current_Pending_Sector  -O--CK   200   200   000    -    2

Serious issue #1:

You have unreadable sectors on sdc.  When you hit them during rebuild,
sdc will be kicked out (again).  They might not be permanent errors, but
you can't tell until the drive is given fresh data to write over them.

You have two choices:

1) use ddrescue to copy sdc onto a new drive, then use it in place of
sdc when you re-create the array, or

2) use badblocks to find the exact locations of the bad sectors, then
write zeros to those sectors using dd.

Either way, you have lost whatever those sectors used to hold.

befor i will recreate the raid with an older mdadm i would search the 
badblocks. is this right?
i have check all drives and the sdc device had badblock:
Pass completed, 48 bad blocks found. (48/0/0 errors)
but die binary dont give me the info where they are..
i have used this command in a screen badblocks -v /dev/sdc1

quoted

quoted

[trim /]

yes this cheep WD Green drives. i have 4 new better drives here the i
will use instead. this means i will get the raid running and than i copy
all the data on the new drives.

quoted

quoted

SCT Status Version:                  3
SCT Version (vendor specific):       258 (0x0102)
SCT Support Level:                   1
Device State:                        Active (0)
Current Temperature:                    36 Celsius
Power Cycle Min/Max Temperature:     33/37 Celsius
Lifetime    Min/Max Temperature:     33/44 Celsius
Under/Over Temperature Limit Count:   0/0
SCT Temperature History Version:     2
Temperature Sampling Period:         1 minute
Temperature Logging Interval:        1 minute
Min/Max recommended Temperature:      0/60 Celsius
Min/Max Temperature Limit:           -41/85 Celsius
Temperature History Size (Index):    478 (314)

Index    Estimated Time   Temperature Celsius
   315    2013-02-19 14:26    36  *****************
   ...    ..(476 skipped).    ..  *****************
   314    2013-02-19 22:23    36  *****************

Warning: device does not support SCT Error Recovery Control command

Serious issue #2:

Error timeout mismatch.  Your cheap drives do not support Error Recovery
Control.  That means when they run into unreadable sectors, they will
spend a couple minutes trying "extra hard" to get the data.

But linux is only going to wait 30 seconds.  Then it will reset the SATA
link and try again.  But the drive will *not* give up its error recovery
effort, and will not even *talk* to the linux driver in the meantime, so
the linux driver will disconnect the drive and report errors for all
remaining requests.  This will cause MD to kick the drive out.

You only have one choice:

1) Set a long timeout in the linux drivers for the drives in your array,
on every boot.  Something like:

for x in /sys/block/sd[cdef]/device/timeout ; do echo 180 >$x ; done

If you had slightly better drives, SCTERC would be supported.  On
desktop drives at power up, it is disabled.  But you would be able to
enable a normal 7.0 second timeout in the drives using smartctl.  (In a
script, on every boot up.)  Enterprise "raid" drives do this by default.

[trim /]

quoted

smartctl -x /dev/sdd1
smartctl 5.41 2011-06-09 r3365 [x86_64-linux-3.2.0-23-generic] (local
build)
Copyright (C) 2002-11 by Bruce Allen,
http://smartmontools.sourceforge.net

[trim /]

quoted

SMART Attributes Data Structure revision number: 16
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME          FLAGS    VALUE WORST THRESH FAIL RAW_VALUE
    1 Raw_Read_Error_Rate     POSR-K   200   200   051    -    534
    3 Spin_Up_Time            POS--K   172   171   021    -    6383
    4 Start_Stop_Count        -O--CK   100   100   000    -    586
    5 Reallocated_Sector_Ct   PO--CK   200   200   140    -    2

You already have two relocations on this drive.

quoted

    7 Seek_Error_Rate         -OSR-K   100   253   000    -    0
    9 Power_On_Hours          -O--CK   085   085   000    -    11487

In less than two years.  You should pay close attention to this.

Phil

i think i must learn to interpret the smart values better.
thank you.
i will send you tomorrow my new info with the older mdadm version.