Topic: hdparm in SME 5.6

[Robbie]

It does not appear that the hdparm service is in SME 5.6. This means that I have to start udma mode manually (hdparm -d 1 /dev/hda) every time I restart my SME 5.6 box to get any type of hard drive performance out of my IDE hard drives.

My question is does anyone know how to start udma mode automatically every time the SME 5.6 box starts. I know that in 5.5 you could run the "/sbin/e-smith/config setprop hdparm status enabled" command and it udma would be enabled from that point on, but this does not work in 5.6.

If anyone knows how to fix this problem I would greatly appreciate the help.

[Charlie Brady]

Robbie wrote:
 
> It does not appear that the hdparm service is in SME 5.6.
> This means that I have to start udma mode manually (hdparm -d
> 1 /dev/hda) every time I restart my SME 5.6 box to get any
> type of hard drive performance out of my IDE hard drives.

Interesting. Mine comes up in udma mode automatically.

> My question is does anyone know how to start udma mode
> automatically every time the SME 5.6 box starts.

Yes, if you want the same setting across all hard drives, edit /etc/sysconfig/harddisks. If you want different settings, copy /etc/sysconfig/harddisks to /etc/sysconfig/harddiskhda and /etc/sysconfig/harddiskhdc (etc) and choose the settings in each. This is exactly the same as what you would do with RedHat.

Regards

Chalrie

[eric]

I never really checked to see if I had this enabled...is there a way I can remotely check?  My server is about 20 miles away and I never get to be there when it's rebooted to watch the startup process...

[Boris]

ssh to you server and type
# hdparm -d /dev/hda
it will tell you if DMA = 1 (on) or 0 (off)
# hdparm -t /dev/hda
 will test the transfer speed. On today's disks it may be somewhere in 20-40s depends on MB and drive. If DMA is OFF it likely to be 3-5 Mbs

[Charlie Brady]

Robbie wrote:

> This means that I have to start udma mode manually (hdparm -d
> 1 /dev/hda) every time I restart my SME 5.6 box to get any
> type of hard drive performance out of my IDE hard drives.

Are you sure that hard drive performance makes any practical difference to the performance of you system? For most server activities, it doesn't make much practical difference. Just a thought.

Charlie

[Robbie]

Thank you Charlie for the information. Your solutions fixed my problem.

Yes, the hard drive performance makes a huge difference to SME server. The main purpose of my SME server is to be a file server. With the udma option disabled the server is extremely slow when transferring large files. When I enable the udma option the server screams.

Once again, thank you for your help Charlie. I appreciate your time and knowledge in helping me resolve this issue.

[Graham Harris]

While we are on the subject of Hard Disk performance , can some learned person comment on the following.....

From dmesg:
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx

How do I go about changing that to 66.

I have ABIT BH-6 with Celeron 300a (@450) and a recent 80gb disk.

My newer computer came with the blue colour ide connector on the mobo and an 80pin cable, the BH-6 did not.


Nothing on idebus in the forum or Technician's Handbook - Release 5.6 or Administrator's Guide - Release 5.6

[Guck Puppy]

> Robbie wrote:
> > This means that I have to start udma mode manually (hdparm -d
> > 1 /dev/hda) every time I restart my SME 5.6 box to get any
> > type of hard drive performance out of my IDE hard drives.

Charlie Brady wrote:
> Are you sure that hard drive performance makes any practical
> difference to the performance of you system? For most server
> activities, it doesn't make much practical difference. Just a
> thought.

I know the dma yea or nay is a frequent discussion, but I have to chip in...

whilst dumping a bunch of CD sized files to the new 5.6 server over a switched fast ethernet connection I was shocked at the results. Watching in xp's task manager Networking tab I would get very thin spikes of 60-70% bandwidth usage followed by much larger gaps of next to none. Also, whilst this was going on the ssh session was useless - simply cursor-upping through the history took seconds per.

I thought I had set the dma on the seagate barracuda V 120GB drives already (but must have been one of the other install attempts) so I turned it on.

Now when transferring CD sized files and lots of little ones the line stays pegged between 70 - 80% bandwidth use - and I can use the shell too!

And I think when Robbie said :
"This means that I have to start udma mode manually (hdparm -d"
he was talking about just the dma (-d flag), not the udma (-X flag).

I'm guessing that based on the info from hdparm -i /dev/hda putting a * next to udma5 that that's the current mode it is in...?

G

[Dean Mumby]

Can anyone think why a brandnew seagate 120GB drive would not be able to use dma

its an intel board with a p4 2.4 and if I try
hdparm -d1 /dev/hda
it says failed
surely all new drives should be able to be configured like this

Dean

[Guck Puppy]

Dean Mumby wrote:

> hdparm -d1 /dev/hda
> it says failed

hdparm -d 1 /dev/hda

Also, to retain this after reboot, remember to edit
/etc/sysconfig/harddisks

G

[Mike]

Used to set hdparm to enabled on my older SME 5.5 box, no problem.
Now I have a new Athlon XP2200 system with a 133-IDE and my harddisk is a WD 200Gb (I also had that harddisk on my 5.5).
I also found out that my hdparm was not enabled, and I could not enable it the same way as on my older SME 5.5 so I searched the forum and found this threat.
So then I tried to enable it the next way:

hdparm -d 1 /dev/hda

I have been logged on to my box as root but when I try to set hdparm like shown above I get the next:

/dev/hda:
setting using_dma to (on)
HDIO_SET_DMA failed: Operation not permitted
using_dma      =    0 (off)

Edited /etc/sysconfig/harddisk to and rebooted.
Looked easy, much explanation in the harddisks-file.
But, no luck.
Root not enough permission?
No, must be something else...

Hope someone has an answer to that one.

[Dean Mumby]

I have experienced this with the Intel "Brownsville2 " motherboard

I believe the problem is with the "old" kernel that is used with sme-5.6
I am aware of security issues with this kernel and would hope that there are plans to upgrade although I cant recall an e-smith version having a kernel update ... may be wrong though.

I dont think there is any other solution other than to downgrade to sme-5.5 or wait. I have recently upgraded a server to a new P4 on sme-5.6 and it is less responsive than the old P3 with 5.1.2. I was hoping to wait until the next release but the end of support for 5.1.2 has forced my hand.

If you do find a solution I would be very interested.

Regards
Dean

[bud]

Yeah, SME 5.6 (RH 7.3) seems to have poor support for new P4 hardware chipsets.  I'm looking into patching the kernel.

[Mike]

Look here: http://forums.contribs.org/index.php?topic=7880.msg29171#msg29171
This will probably give you a headstart.

Mike