> Do you have a particular goal, or were you just making sure you understood?

Both. My goal is to keep my data a safe as possible in case of a 2 and more disk failure
for non permanent failures. During experiments I has problems with a power cable
and I was not able to recover from this situation. (3 Disk ok, 2 that failed after
each other)
I therefor would not like to start the array when not all disk are ok.
I created a script that checks in readonly mode first.

On 2014-11-08 00:22, NeilBrown wrote:

On Fri, 07 Nov 2014 17:43:07 +0100 "P. Gautschi"[off-list ref]
wrote:

quoted

As far as I understand the documentation --assemble --no-degraded should not start a degraded array.
However on my system (kubuntu 14.10)

# mdadm --assemble --no-degraded /dev/md0 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1 /dev/sdf1
mdadm: /dev/md0 has been started with 4 drives (out of 5).

# mdadm --detail /dev/md0
/dev/md0:
          Version : 1.2
    Creation Time : Tue Nov  4 15:26:46 2014
       Raid Level : raid5
       Array Size : 599469328 (571.70 GiB 613.86 GB)
    Used Dev Size : 149867332 (142.92 GiB 153.46 GB)
     Raid Devices : 5
    Total Devices : 4
      Persistence : Superblock is persistent

    Intent Bitmap : Internal

      Update Time : Fri Nov  7 17:22:53 2014
            State : clean, degraded
   Active Devices : 4
Working Devices : 4
   Failed Devices : 0
    Spare Devices : 0

           Layout : left-symmetric
       Chunk Size : 4K

             Name : 0
             UUID : c7465b19:c149b2d1:5b4d88ce:8c6ce432
           Events : 642

      Number   Major   Minor   RaidDevice State
         0       0        0        0      removed
         1       8       33        1      active sync   /dev/sdc1
         2       8       49        2      active sync   /dev/sdd1
         3       8       65        3      active sync   /dev/sde1
         5       8       81        4      active sync   /dev/sdf1

the array IS started when removing one disk, stopping it, reconnecting the disk and then assemble the array.
Is this the supposed behavior?

Yes, that is the correct behaviour, though I admit that it is slightly
unintuitive.

--no-degraded will cause mdadm to refuse to assemble an array which is more
degraded than it was last time it was active.

So if you have an optimal array, stop it, then try to assemble with some
devices missing, then --no-degraded will cause that to fail.

If the array is already degraded, then there doesn't seem much point in
stopping it from assembling.

Do you have a particular goal, or were you just making sure you understood?

Thanks,
NeilBrown