1
I have a headless media server that I’ve been using for several years now. It’s been entirely reliable, but some time ago 2 of the 4 hard drives started running very slowly.
The two slow functioning hard drives are connected to an SIL3124 RAID adapter. The two fast ones are connected directly to the motherboard. All 4 are different hard drives, WD, Samsung, and Toshiba, two 2TB and two 3TB. The drives are not running in a RAID array. I’m using SnapRAID.
While I’d like to point fingers at the RAID/JBOD adapter, the read/write speeds are at 100MB/s just after boot and for as long as I’m using it. It’s only after I let the drives idle for some period of time—and I don't know how long—that the speed drops and I need to reboot again to make it fast.
hdparm output below:
> sudo hdparm -Tt /dev/sdf1
/dev/sdf1:
Timing cached reads: 6032 MB in 2.00 seconds = 3016.41 MB/sec
Timing buffered disk reads: 10 MB in 3.72 seconds = 2.69 MB/sec
> sudo hdparm -Tt /dev/sde1
/dev/sde1:
Timing cached reads: 5592 MB in 2.00 seconds = 2796.49 MB/sec
Timing buffered disk reads: 8 MB in 3.41 seconds = 2.34 MB/sec
> sudo hdparm -Tt /dev/sdd1
/dev/sdd1:
Timing cached reads: 8212 MB in 2.00 seconds = 4107.42 MB/sec
Timing buffered disk reads: 338 MB in 3.01 seconds = 112.40 MB/sec
> sudo hdparm -Tt /dev/sdc1
/dev/sdc1:
Timing cached reads: 8374 MB in 2.00 seconds = 4188.83 MB/sec
Timing buffered disk reads: 536 MB in 3.01 seconds = 178.30 MB/sec
I have them set to spin down after 2 hours using /etc/hdparm.conf
/dev/sdc {
spindown_time = 244
}
/dev/sdd {
spindown_time = 244
}
/dev/sde {
spindown_time = 244
}
/dev/sdf {
spindown_time = 244
}
I wonder if the drives are stuck in some kind of weird mode after spin down but don't know how to test. SMART says that my all of my drives are fine. Does anyone know what the problem might be or how to debug this issue?
Edit: df -i output
Filesystem Inodes IUsed IFree IUse% Mounted on
/dev/sda1 610800 482095 128705 79% /
udev 484830 558 484272 1% /dev
tmpfs 487028 14 487014 1% /tmp
tmpfs 487028 491 486537 1% /run
none 487028 3 487025 1% /run/lock
none 487028 2 487026 1% /run/shm
tmpfs 487028 1 487027 1% /var/log/apt
none 487028 14 487014 1% /var/cache/apt
/dev/sda2 610800 141230 469570 24% /home
/dev/sdc1 715424 14 715410 1% /mnt/toshiba-3tb-1
/dev/sdd1 122101760 2044 122099716 1% /mnt/wd-2tb-1
/dev/sdf1 183148544 131929 183016615 1% /mnt/wd-3tb-1
/dev/sde1 122101760 3935 122097825 1% /mnt/samsung-2tb-2
Can you post the output of
df -i
to your question? Could be an inode issue. – JakeGould – 2015-09-24T01:28:50.927@JakeGould Added
df -i
output. Does it look right? – Yunchi – 2015-09-24T01:45:41.163Thanks for adding that info. Not 100% sure but good to post. – JakeGould – 2015-09-24T01:55:54.937