ICH10 erratic sata speed

[julianweber]

I am running Proxmox v1.8 on my Supermicro x8sax. The system boots off an Intel RAID controller and the Virtual Machines are on an LSI logic SATA RAID controller. These are working fine. My problem is a Seagate 500GB SATA drive that I have plugged into one of the onboard ICH10 SATA ports. I use this drive strictly for backups so it isn't critical but I think the performance is far below normal. Can someone help me diagnose this and possible find out what I have misconfigured.

Here are the pveperf results...

proxmox:~# pveperf
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 954659
HD SIZE: 8.37 GB (/dev/mapper/pve-root)
BUFFERED READS: 62.69 MB/sec
AVERAGE SEEK TIME: 5.74 ms
FSYNCS/SECOND: 129.18
DNS EXT: 120.03 ms
DNS INT: 1.00 ms (home.netweber.org)

Here are a few hdparm results. These were run back to back with the drive unmounted so there should have been no other activity...

proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 16732 MB in 2.00 seconds = 8375.82 MB/sec
Timing buffered disk reads: 74 MB in 4.03 seconds = 18.35 MB/sec
proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 17110 MB in 2.00 seconds = 8565.51 MB/sec
Timing buffered disk reads: 86 MB in 3.03 seconds = 28.35 MB/sec
proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 16726 MB in 2.00 seconds = 8373.23 MB/sec
Timing buffered disk reads: 140 MB in 3.01 seconds = 46.52 MB/sec
proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 16616 MB in 2.00 seconds = 8317.24 MB/sec
Timing buffered disk reads: 144 MB in 3.02 seconds = 47.74 MB/sec
proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 16912 MB in 2.00 seconds = 8466.21 MB/sec
Timing buffered disk reads: 16 MB in 3.01 seconds = 5.31 MB/sec
proxmox:~# hdparm -Tt /dev/sdc

/dev/sdc:
Timing cached reads: 17194 MB in 2.00 seconds = 8607.34 MB/sec
Timing buffered disk reads: 96 MB in 8.85 seconds = 10.85 MB/sec



Thanks for your help,
Julian Weber

 

[udo]

Hi,
what values shows pveperf if you test it like

pveperf /backup

And how is the performance with bonnie++? You can also look with "iostat -dm 5 sdc" during the test (you need the package sysstat).

Some time ago someone has performance-problems with an supermicro-board (AFAIK bios-settings/bios-upgrade) - use the find-function of this forum.

Udo

 

[julianweber]

Hi,
what values shows pveperf if you test it like

pveperf /backup

Udo

Here is the quick answer running pveperf on the backup drive mounted as /mnt/backup. I ran it three times back to back to show the variation.
I will have to figure out how to run bonnie++ for the other part.

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 965905
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 25.01 MB/sec
AVERAGE SEEK TIME: 123.94 ms
FSYNCS/SECOND: 27.11
DNS EXT: 41.22 ms
DNS INT: 1.00 ms (home.netweber.org)

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 969619
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 20.57 MB/sec
AVERAGE SEEK TIME: 20.80 ms
FSYNCS/SECOND: 16.91
DNS EXT: 42.29 ms
DNS INT: 0.99 ms (home.netweber.org)

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 965342
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 30.61 MB/sec
AVERAGE SEEK TIME: 43.17 ms
FSYNCS/SECOND: 19.17
DNS EXT: 39.48 ms
DNS INT: 1.02 ms (home.netweber.org)

 

[udo]

Here is the quick answer running pveperf on the backup drive mounted as /mnt/backup. I ran it three times back to back to show the variation.
I will have to figure out how to run bonnie++ for the other part.

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 965905
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 25.01 MB/sec
AVERAGE SEEK TIME: 123.94 ms
FSYNCS/SECOND: 27.11
DNS EXT: 41.22 ms
DNS INT: 1.00 ms (home.netweber.org)

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 969619
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 20.57 MB/sec
AVERAGE SEEK TIME: 20.80 ms
FSYNCS/SECOND: 16.91
DNS EXT: 42.29 ms
DNS INT: 0.99 ms (home.netweber.org)

proxmox:~# pveperf /mnt/backup/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 965342
HD SIZE: 457.27 GB (/dev/sdc4)
BUFFERED READS: 30.61 MB/sec
AVERAGE SEEK TIME: 43.17 ms
FSYNCS/SECOND: 19.17
DNS EXT: 39.48 ms
DNS INT: 1.02 ms (home.netweber.org)

Hi,
if there are not other (heavy) access on this disk it's a "no go"!

bonnie++ like this:

bonnie++ -u root -f -n 0 -r 8000 -d testdir

but pveperf shows that there is something wrong (if no other processes create disk-load).

Do you have tried other bios-settings? Is the bios up to date?

Are there error-messages in the logs? Perhaps your disk is not very healthy (what shows "smartctl --all /dev/sdc" (package smartmontools)?)

Udo

 

[julianweber]

Hi,
if there are not other (heavy) access on this disk it's a "no go"!
Udo

There was nothing accessing the disk during these tests.

Hi,
Do you have tried other bios-settings? Is the bios up to date?

Are there error-messages in the logs? Perhaps your disk is not very healthy (what shows "smartctl --all /dev/sdc" (package smartmontools)?)

Udo

Here are the logs after mounting...

Jul 21 14:39:46 proxmox kernel: XFS mounting filesystem sdc4
Jul 21 14:39:56 proxmox kernel: ata1: hard resetting link
Jul 21 14:39:56 proxmox kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
Jul 21 14:39:57 proxmox kernel: ata1.00: configured for UDMA/33
Jul 21 14:39:57 proxmox kernel: ata1: EH complete
Jul 21 14:40:00 proxmox kernel: ata1: hard resetting link
Jul 21 14:40:00 proxmox kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
Jul 21 14:40:05 proxmox kernel: ata1.00: qc timeout (cmd 0xec)
Jul 21 14:40:05 proxmox kernel: ata1.00: failed to IDENTIFY (I/O error, err_mask=0x4)
Jul 21 14:40:05 proxmox kernel: ata1: hard resetting link
Jul 21 14:40:06 proxmox kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
Jul 21 14:40:07 proxmox kernel: ata1.00: configured for UDMA/33
Jul 21 14:40:07 proxmox kernel: ata1: EH complete
Jul 21 14:40:20 proxmox kernel: ata1: hard resetting link
Jul 21 14:40:21 proxmox kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
Jul 21 14:40:22 proxmox kernel: ata1.00: configured for UDMA/33
Jul 21 14:40:22 proxmox kernel: ata1: EH complete
Jul 21 14:40:29 proxmox kernel: ata1.00: configured for UDMA/33
Jul 21 14:40:29 proxmox kernel: ata1.00: device reported invalid CHS sector 0
Jul 21 14:40:29 proxmox kernel: ata1: EH complete
Jul 21 14:40:40 proxmox kernel: ata1: hard resetting link
Jul 21 14:40:41 proxmox kernel: ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
Jul 21 14:40:42 proxmox kernel: ata1.00: configured for UDMA/33
Jul 21 14:40:42 proxmox kernel: ata1: EH complete


I am going to try removing this drive from the esata case and installing it internally.

 

[julianweber]

OK. I installed an older wd2500ks 250GB sata drive internally and the results are a bit slow, but I am fairly certain that my old backup drive or caddy is going bad.

CPU BOGOMIPS: 42558.71
REGEX/SECOND: 966191
HD SIZE: 229.23 GB (/dev/sdd1)
BUFFERED READS: 58.00 MB/sec
AVERAGE SEEK TIME: 13.25 ms
FSYNCS/SECOND: 25.49
DNS EXT: 41.84 ms
DNS INT: 5.08 ms (home.netweber.org)
proxmox:~# pveperf /mnt/tmp/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 952922
HD SIZE: 229.23 GB (/dev/sdd1)
BUFFERED READS: 57.92 MB/sec
AVERAGE SEEK TIME: 13.49 ms
FSYNCS/SECOND: 27.63
DNS EXT: 43.00 ms
DNS INT: 0.97 ms (home.netweber.org)
proxmox:~# pveperf /mnt/tmp/
CPU BOGOMIPS: 42558.71
REGEX/SECOND: 942153
HD SIZE: 229.23 GB (/dev/sdd1)
BUFFERED READS: 57.96 MB/sec
AVERAGE SEEK TIME: 13.25 ms
FSYNCS/SECOND: 25.42
DNS EXT: 40.49 ms
DNS INT: 1.01 ms (home.netweber.org)