Slow md RAID-1 array
We have some problems with very slow disk response on our server. I checked iostat (iostat -d -x 30) and have some problems with its interpretation:
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 1.04 396.31 6.60 57.44 382.47 3649.21 62.95 10.31 160.87 8.64 55.36
sda 6.26 391.15 16.16 62.75 1810.79 3649.22 69.19 2.97 37.66 1.79 14.13
md0 0.00 0.00 0.55 0.01 16.88 0.08 30.11 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.02 0.07 1.10 0.54 18.31 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.02 0.04 0.13 0.34 8.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 29.48 453.28 2175.15 3643.46 12.05 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 56.15 0.70 81.34 12.00 1110.03 13.68 47.56 600.17 5.23 42.89
sda 0.00 51.02 0.47 81.37 4.53 1059.38 13.00 0.32 3.95 0.69 5.64
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 1.17 47.45 16.53 379.61 8.15 0.00 0.00 0.00 0.00
First one is initial (history) statistics iostat displays, the second one is after 30 seconds.
Why is await for sdb so higher than for sda? OK, because svctm is also higher (svctm is part of await but also influences the queue length). But why, if there are in mirror? They are both exactly the same disks, smartctl does not report any problems or significant counters difference:
SDA:
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE
1 Raw_Read_Error_Rate 0x002f 200 200 051 Pre-fail Always - 0
3 Spin_Up_Time 0x0027 145 145 021 Pre-fail Always - 9716
4 Start_Stop_Count 0x0032 100 100 000 Old_age Always - 71
5 Reallocated_Sector_Ct 0x0033 200 200 140 Pre-fail Always - 0
7 Seek_Error_Rate 0x002e 200 200 000 Old_age Always - 0
9 Power_On_Hours 0x0032 080 080 000 Old_age Always - 14623
10 Spin_Retry_Count 0x0032 100 253 000 Old_age Always - 0
11 Calibration_Retry_Count 0x0032 100 253 000 Old_age Always - 0
12 Power_Cycle_Count 0x0032 100 100 000 Old_age Always - 69
192 Power-Off_Retract_Count 0x0032 200 200 000 Old_age Always - 68
193 Load_Cycle_Count 0x0032 113 113 000 Old_age Always - 262965
194 Temperature_Celsius 0x0022 126 114 000 Old_age Always - 26
196 Reallocated_Event_Count 0x0032 200 200 000 Old_age Always - 0
197 Current_Pending_Sector 0x0032 200 200 000 Old_age Always - 0
198 Offline_Uncorrectable 0x0030 200 200 000 Old_age Offline - 0
199 UDMA_CRC_Error_Count 0x0032 200 200 000 Old_age Always - 0
200 Multi_Zone_Error_Rate 0x0008 200 200 000 Old_age Offline - 0
SDB:
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE
1 Raw_Read_Error_Rate 0x002f 200 200 051 Pre-fail Always - 0
3 Spin_Up_Time 0x0027 145 145 021 Pre-fail Always - 9708
4 Start_Stop_Count 0x0032 100 100 000 Old_age Always - 67
5 Reallocated_Sector_Ct 0x0033 200 200 140 Pre-fail Always - 0
7 Seek_Error_Rate 0x002e 200 200 000 Old_age Always - 0
9 Power_On_Hours 0x0032 080 080 000 Old_age Always - 14622
10 Spin_Retry_Count 0x0032 100 253 000 Old_age Always - 0
11 Calibration_Retry_Count 0x0032 100 253 000 Old_age Always - 0
12 Power_Cycle_Count 0x0032 100 100 000 Old_age Always - 65
192 Power-Off_Retract_Count 0x0032 200 200 000 Old_age Always - 63
193 Load_Cycle_Count 0x0032 113 113 000 Old_age Always - 261839
194 Temperature_Celsius 0x0022 128 115 000 Old_age Always - 24
196 Reallocated_Event_Count 0x0032 200 200 000 Old_age Always - 0
197 Current_Pending_Sector 0x0032 200 200 000 Old_age Always - 0
198 Offline_Uncorrectable 0x0030 200 200 000 Old_age Offline - 0
199 UDMA_CRC_Error_Count 0x0032 200 200 000 Old_age Always - 0
200 Multi_Zone_Error_Rate 0x0008 200 200 000 Old_age Offline - 0
/etc/fstab:
proc /proc proc defaults 0 0
/dev/md0 / ext3 relatime,errors=remount-ro 0 1
/dev/md1 /var ext3 relatime 0 2
/dev/md2 none swap sw 0 0
/dev/md3 /vz ext3 relatime 0 3
/dev/hda /media/cdrom0 udf,iso9660 user,noauto 0 0
Some measurements with iostat -d -x 2 (every two seconds) under heavy load. You can see that both disks can have longer queue and waiting time, but sda successfully reduces this, while sdb keeps having the longer waiting time. This is strange as disks are the same and it is RAID-1 (mirror).
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 6.00 0.00 6144.00 1024.00 21.40 4545.00 166.67 100.00
sda 0.00 0.00 2.00 1.00 16.00 8.00 8.00 0.49 390.00 75.33 22.60
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 1.50 0.50 12.00 4.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 1405.00 0.50 23.00 4.00 10632.00 452.60 18.96 1889.62 41.62 97.80
sda 0.50 1401.50 1.50 37.50 120.00 11512.00 298.26 4.29 110.00 3.13 12.20
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 2.50 1439.00 124.00 11512.00 8.07 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 1995.50 0.00 29.00 0.00 5304.00 182.90 13.64 873.31 34.34 99.60
sda 0.50 1986.50 6.50 28.50 512.00 1664.00 62.17 0.57 7.14 1.89 6.60
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 7.00 2046.00 512.00 16368.00 8.22 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 930.00 0.00 18.50 0.00 1192.00 64.43 92.52 859.68 54.05 100.00
sda 0.00 928.50 0.00 35.50 0.00 18192.00 512.45 51.52 701.97 28.17 100.00
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.00 946.50 0.00 7572.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 16.00 0.00 8976.00 561.00 56.14 2710.38 62.50 100.00
sda 0.00 0.00 0.00 13.50 0.00 4084.00 302.52 6.26 2457.63 47.56 64.20
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 10.00 0.00 10240.00 1024.00 33.75 4877.20 100.00 100.00
sda 0.00 0.00 0.50 0.00 4.00 0.00 8.00 0.01 16.00 16.00 0.80
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.50 0.00 4.00 0.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 3245.50 1.50 31.50 208.00 12756.00 392.85 64.57 2644.30 30.24 99.80
sda 0.00 3245.00 2.00 60.50 108.00 26444.00 424.83 17.03 272.42 4.61 28.80
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 3.50 3305.50 316.00 26444.00 8.09 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 8.00 0.00 8192.00 1024.00 74.48 2241.50 125.00 100.00
sda 0.00 0.00 0.00 1.00 0.00 8.00 8.00 0.00 0.00 0.00 0.00
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.00 1.00 0.00 8.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 3.00 0.00 22.50 0.00 5192.00 230.76 58.21 3204.18 44.44 100.00
sda 0.00 3.00 3.50 6.50 48.00 76.00 12.40 0.09 8.00 5.60 5.60
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 4.00 10.00 52.00 80.00 9.43 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.50 4098.50 1.50 31.50 324.00 4160.00 135.88 78.08 3401.39 30.24 99.80
sda 0.50 4084.00 2.00 32.00 216.00 8200.00 247.53 57.79 27.53 15.35 52.20
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 4.00 4173.00 536.00 33384.00 8.12 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 20.50 0.00 9228.00 450.15 97.71 1776.78 48.78 100.00
sda 0.00 0.00 0.00 32.00 0.00 13536.00 423.00 72.55 1675.31 31.25 100.00
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 13.00 0.00 7220.00 555.38 67.20 3830.46 76.92 100.00
sda 0.00 0.00 0.00 25.50 0.00 11652.00 456.94 38.91 4491.14 39.22 100.00
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.50 0.50 4.00 4.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 11.00 0.00 5548.00 504.36 50.62 6367.45 90.91 100.00
sda 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.37 0.00 0.00 100.00
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 7.50 0.00 6648.00 886.40 28.48 7513.07 133.33 100.00
sda 0.00 0.00 1.50 3.50 12.00 28.00 8.00 0.24 560.80 20.80 10.40
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 1.50 2.00 12.00 16.00 8.00 0.00 0.00 0.00 0.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz avgqu-sz await svctm %util
sdb 0.00 0.00 0.00 10.00 0.00 4036.00 403.60 12.15 9193.00 100.00 100.00
sda 0.00 0.00 1.00 0.50 8.00 4.00 8.00 0.02 14.67 14.67 2.20
md0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
md3 0.00 0.00 1.00 0.50 8.00 4.00 8.00 0.00 0.00 0.00 0.00
There's an option -W or --write-mostly which is described very similar to what you get: «…
This is valid for RAID1 only and means that the 'md' driver will avoid reading from these devices if at all possible. This can be useful if mirroring over a slow link.
…» — man mdadm
Check it out. This could be the issue.