1 ==================
2 Partial Parity Log
3 ==================
4
5 Partial Parity Log (PPL) is a feature availabl
6 addressed by PPL is that after a dirty shutdow
7 may become inconsistent with data on other mem
8 in degraded state, there is no way to recalcul
9 disks is missing. This can lead to silent data
10 array or using it is as degraded - data calcul
11 that have not been touched by a write request
12 be incorrect. Such condition is known as the R
13 this, md by default does not allow starting a
14
15 Partial parity for a write operation is the XO
16 modified by this write. It is just enough data
17 write hole. XORing partial parity with the mod
18 the stripe, consistent with its state before t
19 which chunk writes have completed. If one of t
20 this stripe is missing, this updated parity ca
21 contents. PPL recovery is also performed when
22 unclean shutdown and all disks are available,
23 the array. Because of this, using write-intent
24 supported.
25
26 When handling a write request PPL writes parti
27 parity are dispatched to disks. PPL is a distr
28 array member drives in the metadata area, on t
29 stripe. It does not require a dedicated journ
30 reduced by up to 30%-40% but it scales with th
31 and the journaling drive does not become a bot
32 failure.
33
34 Unlike raid5-cache, the other solution in md f
35 not a true journal. It does not protect from l
36 silent data corruption. If a dirty disk of a s
37 performed for this stripe (parity is not updat
38 arbitrary data in the written part of a stripe
39 case the behavior is the same as in plain raid
40
41 PPL is available for md version-1 metadata and
42 metadata arrays. It can be enabled using mdadm
43
44 There is a limitation of maximum 64 disks in t
45 keep data structures and implementation simple
46 are not likely due to high risk of multiple di
47 should not be a real life limitation.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.