[Xen-users] Re: [CentOS] SATA disk issue

[Chuck Mattern <camattern@xxxxxxx>]

Following up to myself here, I tracked the problem in my original
posting down to a memory issue with the system board, found that I had
to manually specify the amount of memory using the mem= parameter at
boot time.  Since then Asus has released an updated bios that has
everything working fine without the parameter.  Now there is a new issue
that again appears to be related to the disks.  When attempting to
create a Xen guest I ran into problems with the guest OS filesystem
(created on a logical volume) going read only.  The volume group in
question here is built on two 500 gig SATA disk set for 3Gbps combined
into a RAID0 md.  I traced this down through the dom0 var/log/messages
file to the following error messages:

Nov  7 18:22:43 xen0 kernel: raid0_make_request bug: can't convert block
across chunks or bigger than 256k 427819515 3
Nov  7 18:22:43 xen0 kernel: raid0_make_request bug: can't convert block
across chunks or bigger than 256k 427820029 4
Nov  7 18:22:43 xen0 kernel: raid0_make_request bug: can't convert block
across chunks or bigger than 256k 427820543 3
[deleted numerous duplicate messages]
Nov  7 18:29:53 xen0 kernel: ata1: spurious interrupt (irq_stat 0x8
active_tag -84148995 sactive 0x37fd3fb)

All of the rest of the guests are running on drbd block devices backed
by matching logical volumes on this server and a peer, I created the
same guest on a drbd device as the others are constructed and got three
of the spurious interrupt messages but none of the raid0 messages.

Does this look like a hardware issue or a configuration issue?  My
suspicion is that the buffering of going through drbd slowed the
throughput enough to partially cover for a disk problem,  Any thoughts
on this would be welcomed.

Regards,
Chuck


        I'm in the process of upgrading the hardware in a small cluster
        based on
        CentOS 5 and Xen.  The old hardware had dual ide disks, the new
        dual
        SATA.  In both cases /boot and / are on RAID 1 partitions on
        each of the
        disks.  The remainder of the disk is a RAID 0 with volume group
        laid on
        top.  Here's my problem, when booting to Xen the box gets to the
        point
        of mounting filesystems for dom0, detects the logical volumes
        but cannot
        find an ext3 filesystem on md1 then panics.  If booting straight
        to
        Linux all goes fine.  The same configuration is running fine on
        the box
        with the ide disks.  I'm wondering if the way Xen handles SATA
        (running
        as SCSI devices (sda, sdb, etc.) might be part of this, seems
        unlikely
        as the lv's in the vg are at least detected which would seem to
        be
        comparable to being able to detect an ext3 filesystem.  
        
        FWIW I checked the grub.conf against the old config and there
        are no
        pertinent differences and I compared the initrd's pointed to by
        each
        stanza and they are identical (extracted to sepearate
        directories, did a
        diff on the directories, compared the init scripts from the old
        initrd
        and the new and found nothing not related to the additional
        hardware and
        all seeming to be in order.  Other potentially significant
        differences:
        
        AMD Athlon 4400 dual core CPU vs. Intel 2.4 GHz
        4 gigs RAM vs. 2 gig RAM
        ASUS M2A-VM system board vs. IBM NetVista
        
        Any suggestions would be greatly appreciated.
        
        Regards,
        Chuck


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users