Here's the final on this one. After testing various scenarios with and
without compression etc. and still getting errors we installed a
bootable memtest86 and let it run for the weekend. On Monday we found
uncountable errors reported. Since we had no spare cpu or compatible
machine to test the supposedly faulty memory (to eliminate possible cpu
error etc.), we took a gamble and bought another stick. As it turns out
we were lucky and found that the original memory was at fault. With the
new memory in, we ran the memtest86 again for 24 hours to find there
were no errors. All tars/gzips/amanda backups now work great! Thanks for
all your support and help everyone.
t.
On Fri, 2005-02-18 at 21:12, Gene Heskett wrote:
> On Friday 18 February 2005 11:49, Thomas Charles Robinson wrote:
> >Thanks John for all your input:
> >
> >On Fri, 2005-02-18 at 15:50, Gene Heskett wrote:
> >Looking down below at the errors, I'm wondering if up2date or yum
> > has
> >
> >> replaced your version of tar, known to be good, with another in
> >> the 1.14 family, known to be bad? 1.15 however seems to be fine,
> >> I've been using it for about a month and have yet to see an
> >> amverify error, which my own scripts run after amdump and its
> >> wrappers has exited.
> >
> >$ gzip -V
> >gzip 1.3.3
> >(2002-03-08)
>
> Same as mine
>
> >$ tar --version
> >tar (GNU tar) 1.13.25
>
> Should be good, but its possible the path to tar is hardcoded into
> amanda, mine is, and all I have to do is change that path, recompile
> and reinstall and I'm back on the older 1.13-25. I took advantage of
> that when I thought I'd see if the lastest tar-1.15-1 worked with
> amanda, and it does. So if you kept a copy of the configuration
> driver script (doesn't everybody?), you might want to see if it
> points at the same tar as a 'which tar' does.
>
> >An interesting point is that after a second run of my test 'some' of
> > the dump-files verified as good. This indicates a intermittent
> > problem. Would bad memory gives this type of behaviour?
>
> Yes. It may also indicate a drive cableing problem, but after
> verifying they're all plugged in solidly, I think I'd reboot on a
> recent cd of linux and run the memtest86 utility, say for about 24
> hours, unless you get a definitive bad memory output before that
> time. That util is also available as a downloadable floppy image
> thats self booting once dd'd to the floppy. If memtest86 comes up
> with a clean bill of health, next would be drive cabling, tested by
> replacement with fresh ones.
>
> Another problem might be cpu cooling. If the grease under the
> heatsink and fan has been there more than a year, its more than
> likely you could cool the cpu off by 10-30 degress F, just by pulling
> the cooler back off, blowing that 2 oz of dust bunnies out of it,
> cleaning off the bottom of it and the top of the cpu, and replaceing
> the grease with the appropriate stuff, some white silicone heat sink
> compound, or in real coffee boiler 2ghz and up cpu's, a dab of 'Artic
> Silver' which is 95% silver ground up micron fine, with a bit of
> silicone grease to make a putty like substance out of it, and which
> if properly done, can cut the degrees per watt differential between
> the top pf the cpu and the bottom of the heat sink by half or more.
> It doesn't take very much, half a matchhead evenly spread should do
> it just fine since that roughly 1cm sq area on the top of the cpu is
> all that needs it.
>
> The reason I note that is that allthough I'm running setiathome here,
> which keeps the cpu running at 100%, when gzip -best is running, the
> temps go up another degree or 2. Zalman Flower cooler here, and my
> grease is drying out, its up about 13 degress since it was installed
> last summer. But I usually practice what I preach as all this stuff
> goes out on the front porch to be cleaned out with an air compressor
> and the heat sinks regreased every spring. Currently showing 138-140
> degrees in gkrellm.
>
> [...]
__--__-- Editing is a rewording activity.
signature.asc
Description: This is a digitally signed message part
|
