Adding a component to an array which has changed size
The amount of space reserved in each component for an internal bitmap depends on the size of the component. If I have (say) a raid5 arrray of size 50GB that used to be a raid5 array of size 1GB, and I fail one of the slots and add in a new replacement, the new slot needs to be bigger than all the existing ones because a larger space is reserved by mdadm for the internal bitmap.
3# mdadm --version && uname -sv mdadm - v2.6.7 - 6th June 2008 Linux 2.6.27.7 #2 SMP Wed Dec 3 12:00:03 GMT 2008 3# mdadm --fail /dev/md127 /dev/mapper/slot.51000.1 mdadm: set /dev/mapper/slot.51000.1 faulty in /dev/md127 3# mdadm --remove /dev/md127 /dev/mapper/slot.51000.1 mdadm: hot removed /dev/mapper/slot.51000.1
3# # wipe superblock so this is a new slot rather than a re-add... 3# dd if=/dev/zero of=/dev/mapper/slot.51000.1 bs=1M count=1 1+0 records in
1+0 records out
3# mdadm --add /dev/md127 /dev/mapper/slot.51000.1
mdadm: /dev/mapper/slot.51000.1 not large enough to join array
3# mdadm --examine /dev/mapper/slot.51000.0
/dev/mapper/slot.51000.0:
Magic : a92b4efc
Version : 1.1
Feature Map : 0x0
Array UUID : 61a6b9ac:4b38148f:c0c0837e:1c34783f
Name : 3:shelf.51000
Creation Time : Thu Dec 4 11:21:49 2008
Raid Level : raid1
Raid Devices : 2 Avail Dev Size : 104857584 (50.00 GiB 53.69 GB)
Array Size : 104857584 (50.00 GiB 53.69 GB)
Data Offset : 16 sectors
Super Offset : 0 sectors
State : active
Device UUID : f8c6dcec:add24452:267858ec:617382cd
Update Time : Thu Dec 4 12:07:22 2008
Checksum : 1e90683 - correct
Events : 19
Array Slot : 0 (0, failed)
Array State : U_ 1 failed
Is there any way either to persuade mdadm to take the bitmap size for from the existing components, or ideally to override the bitmap size when creating arrays and adding components so I can just choose a constant size for all our automatically generated arrays, given that the arrays will vary in size over their lifetimes?
Best wishes,
Chris.
Quoted 11 lines · Chris Webb
The amount of space reserved in each component for an internal bitmap depends on the size of the component. If I have (say) a raid5 arrray of size
50GB that used to be a raid5 array of size 1GB, and I fail one of the slots
and add in a new replacement, the new slot needs to be bigger than all the existing ones because a larger space is reserved by mdadm for the internal bitmap.3# mdadm --version && uname -sv mdadm - v2.6.7 - 6th June 2008
Fixed in v2.6.8, commit 2a528478c75b6659188fc2ce0d9543124992fe6c
NeilBrown
Quoted 58 lines
Linux 2.6.27.7 #2 SMP Wed Dec 3 12:00:03 GMT 2008
3# mdadm --fail /dev/md127 /dev/mapper/slot.51000.1 mdadm: set /dev/mapper/slot.51000.1 faulty in /dev/md127 3# mdadm --remove /dev/md127 /dev/mapper/slot.51000.1 mdadm: hot removed /dev/mapper/slot.51000.13# # wipe superblock so this is a new slot rather than a re-add... 3# dd if=/dev/zero of=/dev/mapper/slot.51000.1 bs=1M count=1 1+0 records in
1+0 records out
3# mdadm --add/dev/md127/dev/mapper/slot.51000.1
mdadm:/dev/mapper/slot.51000.1not large enough to join array3# mdadm --examine /dev/mapper/slot.51000.0 /dev/mapper/slot.51000.0: Magic : a92b4efc Version : 1.1 Feature Map : 0x0 Array UUID : 61a6b9ac:4b38148f:c0c0837e:1c34783f Name : 3:shelf.51000 Creation Time : Thu Dec 4 11:21:49 2008 Raid Level : raid1 Raid Devices : 2Avail Dev Size : 104857584 (50.00 GiB 53.69 GB) Array Size : 104857584 (50.00 GiB 53.69 GB) Data Offset : 16 sectors Super Offset : 0 sectors State : active Device UUID : f8c6dcec:add24452:267858ec:617382cd Update Time : Thu Dec 4 12:07:22 2008 Checksum : 1e90683 - correct Events : 19 Array Slot : 0 (0, failed) Array State : U_ 1 failedIs there any way either to persuade mdadm to take the bitmap size for from the
existing components, or ideally to override the bitmap size when creating arrays and adding components so I can just choose a constant size for all our
automatically generated arrays, given that the arrays will vary in size over
their lifetimes?Best wishes,
Chris.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html