1 ================ 2 RAID 4/5/6 cache 3 ================ 4 5 Raid 4/5/6 could include an extra disk for data cache besides normal RAID 6 disks. The role of RAID disks isn't changed with the cache disk. The cache disk 7 caches data to the RAID disks. The cache can be in write-through (supported 8 since 4.4) or write-back mode (supported since 4.10). mdadm (supported since 9 3.4) has a new option '--write-journal' to create array with cache. Please 10 refer to mdadm manual for details. By default (RAID array starts), the cache is 11 in write-through mode. A user can switch it to write-back mode by:: 12 13 echo "write-back" > /sys/block/md0/md/journal_mode 14 15 And switch it back to write-through mode by:: 16 17 echo "write-through" > /sys/block/md0/md/journal_mode 18 19 In both modes, all writes to the array will hit cache disk first. This means 20 the cache disk must be fast and sustainable. 21 22 write-through mode 23 ================== 24 25 This mode mainly fixes the 'write hole' issue. For RAID 4/5/6 array, an unclean 26 shutdown can cause data in some stripes to not be in consistent state, eg, data 27 and parity don't match. The reason is that a stripe write involves several RAID 28 disks and it's possible the writes don't hit all RAID disks yet before the 29 unclean shutdown. We call an array degraded if it has inconsistent data. MD 30 tries to resync the array to bring it back to normal state. But before the 31 resync completes, any system crash will expose the chance of real data 32 corruption in the RAID array. This problem is called 'write hole'. 33 34 The write-through cache will cache all data on cache disk first. After the data 35 is safe on the cache disk, the data will be flushed onto RAID disks. The 36 two-step write will guarantee MD can recover correct data after unclean 37 shutdown even the array is degraded. Thus the cache can close the 'write hole'. 38 39 In write-through mode, MD reports IO completion to upper layer (usually 40 filesystems) after the data is safe on RAID disks, so cache disk failure 41 doesn't cause data loss. Of course cache disk failure means the array is 42 exposed to 'write hole' again. 43 44 In write-through mode, the cache disk isn't required to be big. Several 45 hundreds megabytes are enough. 46 47 write-back mode 48 =============== 49 50 write-back mode fixes the 'write hole' issue too, since all write data is 51 cached on cache disk. But the main goal of 'write-back' cache is to speed up 52 write. If a write crosses all RAID disks of a stripe, we call it full-stripe 53 write. For non-full-stripe writes, MD must read old data before the new parity 54 can be calculated. These synchronous reads hurt write throughput. Some writes 55 which are sequential but not dispatched in the same time will suffer from this 56 overhead too. Write-back cache will aggregate the data and flush the data to 57 RAID disks only after the data becomes a full stripe write. This will 58 completely avoid the overhead, so it's very helpful for some workloads. A 59 typical workload which does sequential write followed by fsync is an example. 60 61 In write-back mode, MD reports IO completion to upper layer (usually 62 filesystems) right after the data hits cache disk. The data is flushed to raid 63 disks later after specific conditions met. So cache disk failure will cause 64 data loss. 65 66 In write-back mode, MD also caches data in memory. The memory cache includes 67 the same data stored on cache disk, so a power loss doesn't cause data loss. 68 The memory cache size has performance impact for the array. It's recommended 69 the size is big. A user can configure the size by:: 70 71 echo "2048" > /sys/block/md0/md/stripe_cache_size 72 73 Too small cache disk will make the write aggregation less efficient in this 74 mode depending on the workloads. It's recommended to use a cache disk with at 75 least several gigabytes size in write-back mode. 76 77 The implementation 78 ================== 79 80 The write-through and write-back cache use the same disk format. The cache disk 81 is organized as a simple write log. The log consists of 'meta data' and 'data' 82 pairs. The meta data describes the data. It also includes checksum and sequence 83 ID for recovery identification. Data can be IO data and parity data. Data is 84 checksummed too. The checksum is stored in the meta data ahead of the data. The 85 checksum is an optimization because MD can write meta and data freely without 86 worry about the order. MD superblock has a field pointed to the valid meta data 87 of log head. 88 89 The log implementation is pretty straightforward. The difficult part is the 90 order in which MD writes data to cache disk and RAID disks. Specifically, in 91 write-through mode, MD calculates parity for IO data, writes both IO data and 92 parity to the log, writes the data and parity to RAID disks after the data and 93 parity is settled down in log and finally the IO is finished. Read just reads 94 from raid disks as usual. 95 96 In write-back mode, MD writes IO data to the log and reports IO completion. The 97 data is also fully cached in memory at that time, which means read must query 98 memory cache. If some conditions are met, MD will flush the data to RAID disks. 99 MD will calculate parity for the data and write parity into the log. After this 100 is finished, MD will write both data and parity into RAID disks, then MD can 101 release the memory cache. The flush conditions could be stripe becomes a full 102 stripe write, free cache disk space is low or free in-kernel memory cache space 103 is low. 104 105 After an unclean shutdown, MD does recovery. MD reads all meta data and data 106 from the log. The sequence ID and checksum will help us detect corrupted meta 107 data and data. If MD finds a stripe with data and valid parities (1 parity for 108 raid4/5 and 2 for raid6), MD will write the data and parities to RAID disks. If 109 parities are incompleted, they are discarded. If part of data is corrupted, 110 they are discarded too. MD then loads valid data and writes them to RAID disks 111 in normal way.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.