Koozali.org: home of the SME Server
Obsolete Releases => SME Server 7.x => Topic started by: tlicht on May 09, 2010, 11:53:32 PM
-
Hi,
I am not 100% certain that it is a bug I stumbled upon, but it looks like it....
I have been using SME server for several years. I've had the v 7.4 for some years now on a reasonably modern mobo with an Athlon64 xp 3000+ cpu. This used around 90W, so I thought of going green and saving some on the bargain.
I got the GA-D510UD mobo(atom) and did just manage to get past the screen asking about language and keyboard layout when the installation program wants me to choose a storage driver. Naturally none of those in the list satisfied the installation program.... This was a real setback. Disappointed, I put together a server with a mobo of some years and two new 1TB Samsungs. Installation went well and I managed to copy the data from my old SME server.
OK, I now had better capacity, but power consumtion was still 90 - 100 W. In a kind of optimistic and desperate fit, I connected the new 1Tb disks to the atom mobo together with DGE503T NIC (the one from the computer I used for installing the server) . Lo and behold, the server started seemingly ok. Pretty soon I noticed that disk transfer speeds were not what you would expect from a gigabyte network. The speed varied somewhere between 3Mb/s and 400Kb/s.
After checking the network cabling and that file transfer between workstations were good I started to study the server logs. I found the following in messages, which seems to be connected to my problem:
...
kernel: raid1: raid set md2 active with 2 out of 2 mirrors
kernel: kjournald starting. Commit interval 5 seconds
kernel: EXT3-fs: mounted filesystem with ordered data mode.
kernel: ata_piix 0000:00:1f.2: version 2.00ac7-RH1
kernel: ata_piix 0000:00:1f.2: MAP [ P0 P2 P1 P3 ]
kernel: ACPI: PCI Interrupt 0000:00:1f.2[B] -> GSI 19 (level, low) -> IRQ 193
kernel: ata: 0x1F0 IDE port busy
kernel: ata: conflict with ide0
kernel: ata: 0x170 IDE port busy
kernel: ata: conflict with ide1
kernel: ata_piix: probe of 0000:00:1f.2 failed with error -16
kernel: inserting floppy driver for 2.6.9-89.0.16.ELsmp
kernel: raid1: raid set md2 active with 2 out of 2 mirrors
kernel: kjournald starting. Commit interval 5 seconds
kernel: EXT3-fs: mounted filesystem with ordered data mode.
kernel: ata_piix 0000:00:1f.2: version 2.00ac7-RH1
kernel: ata_piix 0000:00:1f.2: MAP [ P0 P2 P1 P3 ]
kernel: ACPI: PCI Interrupt 0000:00:1f.2[B] -> GSI 19 (level, low) -> IRQ 193
kernel: ata: 0x1F0 IDE port busy
kernel: ata: conflict with ide0
kernel: ata: 0x170 IDE port busy
kernel: ata: conflict with ide1
kernel: ata_piix: probe of 0000:00:1f.2 failed with error -16
kernel: inserting floppy driver for 2.6.9-89.0.16.ELsmp
kernel: Floppy drive(s): fd0 is unknown type 13 (usb?)
kernel: FDC 0 is a post-1991 82077
.....
The ata_piix is the correct driver for the ICH7 controller chip on the atom mobo, but obviously it is not quite doing it's job in all instances.
Do you have a suggestion how to solve this - maybe another driver? Which and how to make the server use it? Or any other work around.
-
Check:
http://en.wikipedia.org/wiki/Advanced_Host_Controller_Interface
and make sure your BIOS is set correctly. That will solve your problem.
-
Thanx Byte,
Unfortunately all three settings in BIOS (IDE, AHCI and IDE/RAID) give the exact same result. The only difference in the dmesg files are in the second decimal of the bogomips.
So I still imagine that the ata_piix driver is not up to this chip(set) which is the ICH7 also named 82801 GB/GR/GH.
-
So I still imagine that the ata_piix driver is not up to this chip(set) which is the ICH7 also named 82801 GB/GR/GH.
Try SME8 beta.
-
Can I upgrade or do I have to start from scratch?
-
Can I upgrade or do I have to start from scratch?
Yes you should be able to upgrade, if it doesn't work as expect then be sure to raise a bug.
-
I have too much valuable data on the current server to dare to experiment with the beta.... I have considered to set up a test system with 7.4 and upgrade to 8 to see how and if it works, but I am running a little low on SATA disks..... Maybe I'll have to wait till version 8 is released and it is safe to do the upgrade.
Thanx for valuable tips!
-
you can always try with a virtual machine
-
you can always try with a virtual machine
A virtual machine will not help OP to determine whether hardware compatibility is improved with SME8.
-
A virtual machine will not help OP to determine whether hardware compatibility is improved with SME8.
True, but at least it might make either version (7.4 or 8 beta) work on the obstinate atom mobo....
-
Charlie, I meant that OP could clone his real machine (everything but the hundred GB of file) into a virtual one and try an upgrade..
-
I guess that I am the OP.... 8)
Out of curiosity, I upgraded a 'clean' SME server I have configured for a friend, but not yet 'delivered' to version8 beta5. It went well and seemed to work OK on the 760GXK8MC mobo I used. I then moved the disks to the GS-D510 mobo - it worked so-so. There seems to be no driver for the Realtek RTL8111D NIC, so I added a DGE-530T. Transfer to and from the server from other workstations were still poor. They started in the region of 40 Mbytes/s but degraded very quickly (in just 2 - 4 seconds) to a few hundred Kbytes/s or stopped completely. Probaly still due to ata_piix driver.
Then I tried to install the 8 beta from scratch on the atom and it was very slow! formatting the 500Gb disks took almost 3 hours. After phase 1 of the installation everything halted during the reboot.... so obviously SME-server and the atom mobo are still incompatible.....
-
Then I tried to install the 8 beta from scratch on the atom and it was very slow! formatting the 500Gb disks took almost 3 hours.
According to this:
http://www.thehkstyle.com/blog/?p=226
you need to disable ide0 and ide1 probing, and perhaps force use of the ahci driver. Try again to install SME8, but type:
sme ide0=noprobe ide1=noprobe
at the CDROM boot prompt, and make sure that the append string in grub.conf contains "ide0=noprobe ide1=noprobe". Please report your results via the bug tracker.
BTW. it looks as though your hopes of low power consumption might be frustrated:
http://sandeen.net/wordpress/?p=297
-
Wow!!! That is the info I would have needed from the start. Although I did google around for sata problems with this board, I seemed to miss these.
Regarding the power consumption, I got between 27 and 46 W with two Samsung HD103SI attached to the GA-D510UD. With the NforceE4-A939 with Athlon64 3000+ the power consumption is more like 90 - 100 W with the same disks, so the power savings would have been substantial.
I have now reinstalled the 7.4 on my friend's server and will deliver it today, so I think that I'll have to wait till the release of version 8 before I go back fiddling with the GA-D510UD again.
P.S. checking the buffered disk reads with hdparm -t gives around 75 MB/s on the NforceE4-A939 which is not stellar, but I can live with it for now.
Thanx again for really valuable comments/tips.
-
Wow!!! That is the info I would have needed from the start.
I found by searching for:
GA-D510UD linux slow