Hi all, we're trying to install CentOS 5.1 on a Sun Ultra 40. This is an AMD-powered machine and we're using the x86_64 version of CentOS 5.1. The machine is using the NVidia CK804 chipset and has SATA disks. It also has 16GB's of memory which prompted us to upgrade the BIOS on the machine from 1.1 to 1.6 per this (we also have the two quadro cards):
http://docs.sun.com/source/819-3954-18/index.html#0_37092
The install goes fine up until the installer is trying to format the disks. Part of the way through it simply dies and pops up an error saying that the installer couldn't format the LVM volume and we have to reboot.
An examination of dmesg output shows that there are many SATA errors occuring at this point. Timeouts and such.
After a reboot, the SATA drive no longer shows up -- not even in BIOS. It's as if the formatting has instructed the drive to deactivate itself. :) A hard reset and reseat of the drive in the SATA enclosure brings it back again.
First thought was that the slot or SATA port was bad, so we have moved to others with the same result.
Solaris 10 x86 installs perfectly on this machine, so I'm starting to think that the sata_nv driver is to blame here.
We're in the process of trying 32-bit CentOS 5.1 on the system just for giggles, and may try Fedora 8 as well or RHEL 5.1 and use our paid support to track this issue down, but thought I'd run it by everyone here.
Didn't see any existing issues in bugzilla.redhat.com or bugs.centos.org.
Any insights on this?
I will get the exact error messages posted up here soon (output from dmesg, etc).
TIA, Ray
Ray Van Dolson wrote:
Hi all, we're trying to install CentOS 5.1 on a Sun Ultra 40. This is an AMD-powered machine and we're using the x86_64 version of CentOS 5.1. The machine is using the NVidia CK804 chipset and has SATA disks. It also has 16GB's of memory which prompted us to upgrade the BIOS on the machine from 1.1 to 1.6 per this (we also have the two quadro cards):
http://docs.sun.com/source/819-3954-18/index.html#0_37092
The install goes fine up until the installer is trying to format the disks. Part of the way through it simply dies and pops up an error saying that the installer couldn't format the LVM volume and we have to reboot.
An examination of dmesg output shows that there are many SATA errors occuring at this point. Timeouts and such.
After a reboot, the SATA drive no longer shows up -- not even in BIOS. It's as if the formatting has instructed the drive to deactivate itself. :) A hard reset and reseat of the drive in the SATA enclosure brings it back again.
First thought was that the slot or SATA port was bad, so we have moved to others with the same result.
Solaris 10 x86 installs perfectly on this machine, so I'm starting to think that the sata_nv driver is to blame here.
We're in the process of trying 32-bit CentOS 5.1 on the system just for giggles, and may try Fedora 8 as well or RHEL 5.1 and use our paid support to track this issue down, but thought I'd run it by everyone here.
Didn't see any existing issues in bugzilla.redhat.com or bugs.centos.org.
Any insights on this?
I will get the exact error messages posted up here soon (output from dmesg, etc).
TIA, Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Try booting while disabling ACPI and APIC.
linux noapic noacpi
at the boot menu.
Try booting while disabling ACPI and APIC.
linux noapic noacpi
at the boot menu.
We'll give that a try.
In addition, error message output:
An error occurred trying to format VolGroup00/LogVol00. This problem is serious, and the install cannot continue.
Press <Enter> to reboot your system.
dmesg output <6>sd 0:0:0:0 SCSI error: return code = 0x00040000 <4>end_request: I/0 error, dev sda, sector 504029 <4>printk: 646939 messages suppressed. <3>Buffer I/) error on device dm-0, logical block 19596153 <4>lost page write due to I/O error on dm-0
Again, it sure looks like a hardware problem, but using x86 Solaris 10 on the same disk works perfectly.
Ray
Ray Van Dolson wrote:
Try booting while disabling ACPI and APIC.
linux noapic noacpi
at the boot menu.
We'll give that a try.
In addition, error message output:
An error occurred trying to format VolGroup00/LogVol00. This problem is serious, and the install cannot continue.
Press <Enter> to reboot your system.
dmesg output <6>sd 0:0:0:0 SCSI error: return code = 0x00040000 <4>end_request: I/0 error, dev sda, sector 504029 <4>printk: 646939 messages suppressed. <3>Buffer I/) error on device dm-0, logical block 19596153 <4>lost page write due to I/O error on dm-0
Again, it sure looks like a hardware problem, but using x86 Solaris 10 on the same disk works perfectly.
Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I have seen errors such as this too with some cheap no name software RAID cards. I've also seen it on high end SCSI drives as they do the bad block remapping. I'm not saying it's not hardware, especially given that it specifically mentions sda sector 504029, but it's worth a shot. Solaris might not be reporting the disk remappings or some such thing, who knows. Perhaps it's only doing a quick format and not a full format therefore not even displaying the messages until content begins to be written.
FYI, no difference with noacpi and noapic. Errors this time indicated a different sector, but we'll try with a different SATA disk just to eliminate that as a possibility.
Ray
Ray Van Dolson wrote:
Hi all, we're trying to install CentOS 5.1 on a Sun Ultra 40. This is an AMD-powered machine and we're using the x86_64 version of CentOS 5.1. The machine is using the NVidia CK804 chipset and has SATA disks. It also has 16GB's of memory which prompted us to upgrade the BIOS on the machine from 1.1 to 1.6 per this (we also have the two quadro cards):
http://docs.sun.com/source/819-3954-18/index.html#0_37092
The install goes fine up until the installer is trying to format the disks. Part of the way through it simply dies and pops up an error saying that the installer couldn't format the LVM volume and we have to reboot.
An examination of dmesg output shows that there are many SATA errors occuring at this point. Timeouts and such.
After a reboot, the SATA drive no longer shows up -- not even in BIOS. It's as if the formatting has instructed the drive to deactivate itself. :) A hard reset and reseat of the drive in the SATA enclosure brings it back again.
First thought was that the slot or SATA port was bad, so we have moved to others with the same result.
Solaris 10 x86 installs perfectly on this machine, so I'm starting to think that the sata_nv driver is to blame here.
We're in the process of trying 32-bit CentOS 5.1 on the system just for giggles, and may try Fedora 8 as well or RHEL 5.1 and use our paid support to track this issue down, but thought I'd run it by everyone here.
Didn't see any existing issues in bugzilla.redhat.com or bugs.centos.org.
Any insights on this?
I will get the exact error messages posted up here soon (output from dmesg, etc).
Try "acpi=noirq" as a kernel argument. Some AMD chipsets have problems letting the OS know what irq the 8250 timer is on, the nvidia one is definitely a problem, I have the same chipset in a couple of Dell Dimension e521 desktops :-(
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
Ross S. W. Walker wrote:
Ray Van Dolson wrote:
Hi all, we're trying to install CentOS 5.1 on a Sun Ultra
- This is
an AMD-powered machine and we're using the x86_64 version of CentOS 5.1. The machine is using the NVidia CK804 chipset and has SATA disks. It also has 16GB's of memory which prompted us to upgrade
the BIOS on
the machine from 1.1 to 1.6 per this (we also have the two quadro cards):
http://docs.sun.com/source/819-3954-18/index.html#0_37092
The install goes fine up until the installer is trying to format the disks. Part of the way through it simply dies and pops up an error saying that the installer couldn't format the LVM volume and we have to reboot.
An examination of dmesg output shows that there are many SATA errors occuring at this point. Timeouts and such.
After a reboot, the SATA drive no longer shows up -- not
even in BIOS.
It's as if the formatting has instructed the drive to deactivate itself. :) A hard reset and reseat of the drive in the
SATA enclosure
brings it back again.
First thought was that the slot or SATA port was bad, so we
have moved
to others with the same result.
Solaris 10 x86 installs perfectly on this machine, so I'm
starting to
think that the sata_nv driver is to blame here.
We're in the process of trying 32-bit CentOS 5.1 on the system just for giggles, and may try Fedora 8 as well or RHEL 5.1 and use our paid support to track this issue down, but thought I'd run it by everyone here.
Didn't see any existing issues in bugzilla.redhat.com or bugs.centos.org.
Any insights on this?
I will get the exact error messages posted up here soon (output from dmesg, etc).
Try "acpi=noirq" as a kernel argument. Some AMD chipsets have problems letting the OS know what irq the 8250 timer is on, the nvidia one is definitely a problem, I have the same chipset in a couple of Dell Dimension e521 desktops :-(
My explaination wasn't totally accurate. The acpi=noirq disables the ACPI IRQ routing table lookup for IRQ redirects and reprogramming. Some AMD chipsets had a bug in the way this table was built that caused 2.6 kernels to fail in getting a hook into the table which caused all kinds of intermittent problems. By disabling this feature you run the possibility of IRQ conflicts that will need to use the IRQ management in the BIOS to resolve. Updating the BIOS of the system sometimes fixes the problem.
It just turns out that the system timer irq was my "symptom" that I experienced, but it is different for different systems/configurations.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
Try "acpi=noirq" as a kernel argument. Some AMD chipsets have problems letting the OS know what irq the 8250 timer is on, the nvidia one is definitely a problem, I have the same chipset in a couple of Dell Dimension e521 desktops :-(
My explaination wasn't totally accurate. The acpi=noirq disables the ACPI IRQ routing table lookup for IRQ redirects and reprogramming. Some AMD chipsets had a bug in the way this table was built that caused 2.6 kernels to fail in getting a hook into the table which caused all kinds of intermittent problems. By disabling this feature you run the possibility of IRQ conflicts that will need to use the IRQ management in the BIOS to resolve. Updating the BIOS of the system sometimes fixes the problem.
It just turns out that the system timer irq was my "symptom" that I experienced, but it is different for different systems/configurations.
Thanks Ross. This does indeed make some sense. Currently we're trying with another known-good SATA disk just to rule that out then we'll give this a shot as well.
Ray
Try "acpi=noirq" as a kernel argument. Some AMD chipsets have problems letting the OS know what irq the 8250 timer is on, the nvidia one is definitely a problem, I have the same chipset in a couple of Dell Dimension e521 desktops :-(
Alright, this didn't solve the issue either. So far:
- noacpi noapic (No effect) - New SATA drive (No effect) - acpi=noirq (No effect) - BIOS update to 1.6 (No effect)
Grabbing a copy of RHEL 5.1 now and we'll throw this over to RH support I guess and file something in bugzilla over there.
Gotta be something funky in sata_nv.
Thanks everyone for your responses.
Ray
Hi all, for anyone interested in following this -- ended up having the same issue with RHEL 5.1 and opened a bug (and support request):
https://bugzilla.redhat.com/show_bug.cgi?id=420361
Ray
Ray Van Dolson wrote:
Hi all, for anyone interested in following this -- ended up having the same issue with RHEL 5.1 and opened a bug (and support request):
https://bugzilla.redhat.com/show_bug.cgi?id=420361
Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I was able to use the CentOS 5.0 images to install on the 6 SUN Ultra 40 workstations that I have here in my labs. This allowed me to get around the 5.1 issues, and I performed a yum upgrade to get them to 5.1
I would be interested in further details regarding the fix though if you could please. :)
On Thu, Dec 13, 2007 at 03:05:47PM -0800, James A. Peltier wrote:
Ray Van Dolson wrote:
Hi all, for anyone interested in following this -- ended up having the same issue with RHEL 5.1 and opened a bug (and support request):
https://bugzilla.redhat.com/show_bug.cgi?id=420361
Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I was able to use the CentOS 5.0 images to install on the 6 SUN Ultra 40 workstations that I have here in my labs. This allowed me to get around the 5.1 issues, and I performed a yum upgrade to get them to 5.1
I would be interested in further details regarding the fix though if you could please. :)
James are your machines Ultra 40 M2's or just plain ol Ultra 40's? Ours is the non-M2 model and we had problems with 5.0 as well as 5.1.
However, it appears to be something related to LVM -- specifically something the GUI installer is telling LVM to do. If we manually set up the partitions and LVM stuff (from the console while still at the welcome screen), everything works fine. It's only when anaconda handles the partitioning and LVM setup that we run into issues.
Interesting.
Ray
Ray Van Dolson wrote:
James are your machines Ultra 40 M2's or just plain ol Ultra 40's? Ours is the non-M2 model and we had problems with 5.0 as well as 5.1.
However, it appears to be something related to LVM -- specifically something the GUI installer is telling LVM to do. If we manually set up the partitions and LVM stuff (from the console while still at the welcome screen), everything works fine. It's only when anaconda handles the partitioning and LVM setup that we run into issues.
Interesting.
Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
They're regular ol' SUN Ultra 40s. Running BIOS version 1.400 (I believe) nothing special 2 x Dual Core Opteron 240s running at 2.4GHz w/8GB of RAM.
When trying to install, my experience nearly mirrors yours perfectly. Everything is all fine and dandy until it gets about 50% way through formatting and then throws the ATA errors. When booting rescue with 5.1 I am able to format the already created LVM volume without issue, however it fails with the same errors mentioned when kickstarted.
I just performed an install today on a SUN Ultra 40 and the 5.0 trick worked without issue. Perhaps something in your BIOS? Try loading the default settings and trying again?
So it initially appeared that manually setting up the partitions and LVM seemed to resolve the problem. However, at the time we had swapped out to a different (smaller) hard disk.
When we returned to the original 250GB Sun provided disk, even manually configuring the partitions, LVM and running mkfs.ext3 caused the port to "shut off" and we could not complete the installation.
On a whim, we went into BIOS and disabled LBA for the drive. Lo and behold everyhing worked perfectly during the installation -- fully automated from anaconda.
However -- the system wouldn't boot after the install completed successfully.
To resolve this we had to go into BIOS and re-add the hard disk to the list of bootable devices (weird that it had been removed) and then RE-enable LBA (it will not boot with LBA disabled, doesn't find the boot loader at all).
Now the system boots.
So the trick appears to be:
1. Disable LBA 2. Install 3. Re-enable LBA 4. Ensure drive is in boot device list
Ray
-
James A. Peltier -
Ray Van Dolson -
Ross S. W. Walker