support quality (Re: dump | restore fails: unknown tape headertype 1853384566)

[Peter Jeremy]

On 2009-Mar-25 07:21:37 -0500, "Jack L. Stone" <jacks at sage-american.com> wrote:
>However, last month upon upon upgrading those servers from FBSD-6.3px
>(RELEASE) to 7.0px (RELEASE) we found that about one-half of the servers
>had a similar problem as the original poster while the other half did not.
...
>the next day, and the next day, and the next day. (The servers that did
>work fooled us and we found out about this issue on the others when the
>overlaps appeared and drew our attention). That's when our work to try and
>solve the issues started and went on for days.

It's not clear to me whether all your servers have the problem and you
only initially noticed it on some of them or some of your servers work
and others dont.  In the latter case, you are probably in a very good
position to identify the problem since it is related to some difference
between your servers.

>We could only conclude that the problem was perhaps something with
>hardeware, perhaps the way memory was handled in 7.0, but that is only a
>guess.

If you are talking about server-grade hardware (ECC RAM etc) then it's
unlikely to be RAM corruption.  About the only thing I can think of
would be that if you have RAM above 4GB, you might be running foul of
an address being truncated somewhere (particularly in a device
driver).  (The amd64 user memory map changed between 7.x and 8.x but I
don't think there was any change between 6.x & 7.x).

The pre-emption changes in 7.x and/or moving to an SMP host would seem
to increase the probability of hitting the problem fixed in the patch
mentioned later (kern/121684).

-- 
Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20090326/29a59832/attachment.pgp