I have an issue with an archival server that is running a RAID 5. The server is being accessed only every couple of days, so I want these disks to spin down when there is no activity for a while.
Disclaimer: I understand that spinning down disks is normally bad practice. I am not asking for advice on disk lifespan. I am asking for help on how to make spin downs happen. Thank you.
The file system is ext4. I have ramped up the commit interval of ext4 through the appropriate mount option and verified that there is no activity from jbd2. I have also configured systemd-journald to volatile mode and disabled any other non-essential logging. I have 100% verified that no log files are being written and no user-space processes have IO activity. Swapping is off.
Still, iosnoop is showing periodic writes to sectors 2056, 2064, and 2088 of the disks in the array. I suspect that this is where the superblock or related information is stored. My working theory is that mdadm is marking the RAID as synced or something like that, but I failed to Google any relevant information.
Does anyone have an alternative theory or an idea on how I can stop the IO?
Here’s an iosnoop trace for the first disk in the array:
# iosnoop-perf -s -d "8,16" Tracing block I/O. Ctrl-C to end. STARTs COMM PID TYPE DEV BLOCK BYTES LATms 5068.962692 md0_raid5 249 FF 8,16 18446744073709551615 0 0.35 5068.963054 <idle> 0 WFS 8,16 2064 4096 21.28 5068.990201 md0_raid5 249 FF 8,16 18446744073709551615 0 0.40 5068.990619 <idle> 0 WFS 8,16 2056 512 18.70 5069.017432 kworker/1:1H 216 FF 8,16 18446744073709551615 0 0.42 5069.017866 <idle> 0 WFS 8,16 2088 3072 24.86 5069.442687 md0_raid5 249 FF 8,16 18446744073709551615 0 0.40 5069.443104 <idle> 0 WFS 8,16 2064 4096 7.90 5069.467942 md0_raid5 249 FF 8,16 18446744073709551615 0 0.40 5069.468360 <idle> 0 WFS 8,16 2056 512 57.62 5074.578771 md0_raid5 249 FF 8,16 18446744073709551615 0 0.41 5074.579195 <idle> 0 WFS 8,16 2064 4096 21.82 5084.818728 md0_raid5 249 FF 8,16 18446744073709551615 0 0.41 5084.819146 <idle> 0 WFS 8,16 2088 3072 31.92 5125.794841 md0_raid5 249 FF 8,16 18446744073709551615 0 0.35 5125.795205 <idle> 0 WFS 8,16 2064 4096 22.49 5125.823437 md0_raid5 249 FF 8,16 18446744073709551615 0 0.41 5125.823855 <idle> 0 WFS 8,16 2056 512 18.83 5125.850640 kworker/1:1H 216 FF 8,16 18446744073709551615 0 0.42 5125.851071 <idle> 0 WFS 8,16 2080 4096 8.33 5125.859599 kworker/1:1H 216 FF 8,16 18446744073709551615 0 0.42 5125.860026 <idle> 0 WFS 8,16 2064 4096 7.67 5126.146833 md0_raid5 249 FF 8,16 18446744073709551615 0 3.50 5126.150353 <idle> 0 WFS 8,16 2064 4096 8.98 5126.159498 md0_raid5 249 FF 8,16 18446744073709551615 0 4.39 5126.163913 <idle> 0 WFS 8,16 2056 512 53.75 5131.410989 md0_raid5 249 FF 8,16 18446744073709551615 0 0.41 5131.411412 <idle> 0 WFS 8,16 2064 4096 22.99 5141.650858 md0_raid5 249 FF 8,16 18446744073709551615 0 0.41 5141.651276 <idle> 0 WFS 8,16 2064 4096 16.40 5141.667708 <idle> 0 FF 8,16 18446744073709551615 0 0.29 5141.668012 <idle> 0 WFS 8,16 2080 4096 7.95
Answer
Attribution
Source : Link , Question Author : Philipp Unterbrunner , Answer Author : Community
