Dear Mr. Kabelac,
I have (at first) not been aware, that at the end of your mail, you asked
me probably 'about pseudo-raid'. It is difficult for me to give you an
abstract definition of it, so I give you an example:
* splitting the device '8:16' == /dev/sdb into 17 chunks of equal size by
the following table:
* "dmsetup table ..." --> extended by the mappig 'from the names' to
/dev/dm-<n> (omitting "dmsetup ls", --> mapping "name -> /dev/dm-<n>"
in a separate list).
/dev/dm-20 <-- AFE37B55.A0: 0 86183752 linear 8:16 8
/dev/dm-21 <-- AFE37B55.A1: 0 86183752 linear 8:16 86183760
/dev/dm-22 <-- AFE37B55.A2: 0 86183752 linear 8:16 172367512
/dev/dm-23 <-- AFE37B55.A3: 0 86183752 linear 8:16 258551264
/dev/dm-24 <-- AFE37B55.A4: 0 86183752 linear 8:16 344735016
/dev/dm-25 <-- AFE37B55.A5: 0 86183752 linear 8:16 430918768
/dev/dm-26 <-- AFE37B55.A6: 0 86183752 linear 8:16 517102520
/dev/dm-27 <-- AFE37B55.A7: 0 86183752 linear 8:16 603286272
/dev/dm-28 <-- AFE37B55.A8: 0 86183752 linear 8:16 689470024
/dev/dm-29 <-- AFE37B55.A9: 0 86183752 linear 8:16 775653776
/dev/dm-30 <-- AFE37B55.A10: 0 86183752 linear 8:16 861837528
/dev/dm-31 <-- AFE37B55.A11: 0 86183752 linear 8:16 948021280
/dev/dm-32 <-- AFE37B55.A12: 0 86183752 linear 8:16 1034205032
/dev/dm-33 <-- AFE37B55.A13: 0 86183752 linear 8:16 1120388784
/dev/dm-34 <-- AFE37B55.A14: 0 86183752 linear 8:16 1206572536
/dev/dm-35 <-- AFE37B55.A15: 0 86183752 linear 8:16 1292756288
/dev/dm-36 <-- AFE37B55.A16: 0 86183752 linear 8:16 1378940040
* having a raid4 with these chunks (see /proc/mdstat):
md255 : active raid4 dm-36[17] dm-35[15] dm-34[14] dm-33[13] \
dm-32[12] dm-31[11] dm-30[10] dm-29[9] dm-28[8] \
dm-27[7] dm-26[6] dm-25[5] dm-24[4] dm-23[3] \
dm-22[2] dm-21[1] dm-20[0]
689438720 blocks super 1.0 level 4, 4096k chunk, \
algorithm 0 [17/17] [UUUUUUUUUUUUUUUUU]
(Ok. its actually "raid4", because distributing reads / writes on diffe-
rent legs is not sensible, if these legs reside on the same disk.)
* according to them man page "md": MD(4)
[...]
In later kernels, a read-error will instead
cause md to attempt a recovery by overwriting the bad block. i.e. it
will find the correct data from elsewhere, write it over the block that
failed, and then try to read it back again.
[Modern hard disks have spare sectors; a defective sector is at least,
when it is written to,(sorry for my english) replaced by such a spare-
sector.]
[...]
Sincerely
Thomas
============================================================================
On Mon, November 18, 2024 10:28 am, Zdenek Kabelac wrote:
> Dne 17. 11. 24 v 11:26 "Thomas Bruecker" napsal(a):
>> [...]
> [...]
>> [...]
>> * My name: "pseudo-raid": raid5 with legs on the same device, especially
>> a hard disk. Ok, I admit, its slow.
>> * Reason:
>> * Modern hard disks are able to correct defective sectors, if the
>> driver ('md') knows, what should be in these sectors.
>> * You may not equip a notebook with additional hard disk.
> Is this some sort of 'raid1 + writemostly' ?
> And there is also another target 'dm-era' - which is however not
> supported (yet) by lvm2...
