I decided to check out xfs by using the Redhat 7.1 installer. The
desktop I decided to rebuild has been running as a general purpose
productivity/development workstation for many months running first
RedHat 6.1 then 6.2 with various versions of the 2.2. kernel (most
recently 2.2.19 final).
I rebuilt the machine from scratch using the XGI XFS RedHat 7.1
installer iso burned to a CDRW. The installation went fine. However,
there are some issues. Clearly, there have been a lot of variables
changed (to say the least), and many or all may not be related to xfs at
all. I'm hoping that some of these symptoms may look familiar to some
and help to shortcut the diagnostic steps necesary to attempt to isolate
the problem.
Issue 1: See the following erros in syslog rather frequently:
Apr 22 19:40:33 steve kernel: APIC error on CPU0: 08(01)
Apr 22 19:40:47 steve kernel: APIC error on CPU0: 01(01)
Apr 22 19:40:47 steve kernel: APIC error on CPU1: 02(02)
Apr 22 19:40:51 steve kernel: APIC error on CPU0: 01(01)
Apr 22 19:40:51 steve kernel: APIC error on CPU1: 02(02)
Apr 22 19:40:52 steve kernel: APIC error on CPU1: 02(02)
Apr 22 19:40:52 steve kernel: APIC error on CPU0: 01(01)
Others have reported this in 2.4.x (non-xfs). The response is that the
2.2 kernel was having the same issue, but didn't log it to syslog. The
suggestion is that it might indicate a hardware problem. However this
machine has only locked once in the last year running kernel 2.2. This
is definitely a 2.4 kernel issue; I present it as it may relate to one
or more of these other symptoms.
Issue 2: Seemingly random hard locks, approx once every 8 hours (very
variable). The KDE desktop freezes (the mouse cursor disappears). The
machine does not respond to pings from other hosts on the local net.
Issue 3: After several hours, I see the following symptoms:
top: syslogd is consuming 50-100% of a single processor (this is a
dual-proc pIII)
top: CPU0 is spending 99.7-100.1% of its time in system code
top: CPU1 appears to be normally loaded
I can no longer get a root prompt via su (limiting my ability to debug).
When programs exit, the KDE window manager can't always remove the
window.
Most user space programs continue to run normally.
I know that I'll need to start decomposing the problem by making changes
one at a time, rather than the wholesale change I just went through.
Any thoughts that might help to speed the diag. process would be helpful.
Machine specs. The hardware is unchanged from the relative stability of
2.2 to the issues present in 2.4/xfs.
*Please note that the system's partition table was configured as follows:
P1 : ~ 15MB (unused in the 2.4/xfs config)
P2 : ~ 4GB (Windows 98SE)
P3 : rest of disk (root partition for RH7.1; entire system under 1
partition)
Tyan Tiger 100 dual processor Slot1 motherboard
2 x 700Mhz Intel PIII processors
2 x 128MB PC133 cas2 RAM
Maxtor 20GB EIDE hard drive
ATAPI 100MB internal ZIP drive
Adaptec 2940UW SCSI card w/ CD-ROM and CD-RW attached
Sandisk USB CF reader
Firewire PCI card (TI chipset)
Video capture/TV card (BT848 based)
ISA shutdgun card (2 56k modems)
Creative SBLive! sound card
Hercules Prophet 3D (Nvidia GTS2) video card
Thx for your consideration. If I learn more that suggests an issue with
xfs or xfs interaction with RedHat 7.1, I'll post again to this list.
Steve
|