my situation:
Supermicro H8SCM-F (AMD SR5650+SP5100)
PSU: redundant: 2x48A@12V (this is definitely not the problem)
Ubuntu 10.04.2 x86_64 server
SW-RAID6 over 6 devices
(multiple) SW-RAID1 over 2 devices
onboard AHCI + sata_mv (+ DVD @ sata_sil)
problems only on high IO-load
reproducable with _all_ schedulers (noop,anticipatory,cfq,deadline)
reproducable with kernel options: noapic + acpi=off
I suggest this is a problem in libata layer.
We have some more similar linux boxes with much more drives and SW-RAID6 but they are driven by SAS-controllers.
This problem is really important as our box is a storage server.
I have the very same problem with an opensuse 11.2. (tested once)
Any idea?
My box is ready for tests for the next days/weeks. Any suggestions?
Hi,
I also experience this problem.
my situation: ory,cfq, deadline)
Supermicro H8SCM-F (AMD SR5650+SP5100)
PSU: redundant: 2x48A@12V (this is definitely not the problem)
Ubuntu 10.04.2 x86_64 server
SW-RAID6 over 6 devices
(multiple) SW-RAID1 over 2 devices
onboard AHCI + sata_mv (+ DVD @ sata_sil)
problems only on high IO-load
reproducable with _all_ schedulers (noop,anticipat
reproducable with kernel options: noapic + acpi=off
I suggest this is a problem in libata layer.
We have some more similar linux boxes with much more drives and SW-RAID6 but they are driven by SAS-controllers.
This problem is really important as our box is a storage server.
I have the very same problem with an opensuse 11.2. (tested once)
Any idea?
My box is ready for tests for the next days/weeks. Any suggestions?
Thanks