linux/alpha 2.0.30 - high uptime = hosed machine?

Harvey J. Stein (hjstein@bfr.co.il)
Tue, 23 Jun 1998 17:47:06 +0300


I've been seeing some occasionally strange behavior under linux/alpha
(kernel 2.0.30, rh 4.2 + misc patches).

1. Sometimes update seems to start sucking up lots of CPU:

USER PID %CPU %MEM SIZE RSS TTY STAT START TIME COMMAND
root 4372 99.9 0.2 1960 360 ? S 23:22 14:30 update (bdflush)

2. After a system has been running for a long time, it seems like
hdparm reports much slower cache read times than after a reboot:

Before reboot:

blaster# /sbin/hdparm -Tt /dev/sda

/dev/sda:
Timing buffer-cache reads: 32 MB in 0.53 seconds =60.38 MB/sec
Timing buffered disk reads: 16 MB in 1.22 seconds =13.11 MB/sec

After reboot:

blaster# /sbin/hdparm -Tt /dev/sda

/dev/sda:
Timing buffer-cache reads: 32 MB in 0.26 seconds =121.43 MB/sec
Timing buffered disk reads: 16 MB in 1.18 seconds =13.54 MB/sec

Performance seems to be generally down.

3. After running for a long time, when running a couple of big jobs
the machine seized up. From the noise the disk was making, it
seemed like it was trying to swap to free memory, but was unable.
Shift-scroll-lock was showing 0kb free pages. I forget what it
showed for the number of CLEAN/LOCKED/DIRTY buffers. I eventually
had to power cycle the machine. It wouldn't respond to
cntrl-alt-del, programs were seg faulting all over the place, ...

Is there a memory problem causing such things?

I heard something about memory fragmentation problems. Is this a
symptom? Has it been fixed/improved in the 2.0.34? Has it been
fixed/improved in the 2.1 series?

Thanks,

-- 
Harvey J. Stein
BFM Financial Research
hjstein@bfr.co.il

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu