volume management
Eric Anderson
anderson at freebsd.org
Fri Apr 13 12:58:19 UTC 2007
On 04/12/07 17:34, Ivan Voras wrote:
> Eric Anderson wrote:
>
>> Personally, what I would want to prevent, is having a server go down due
>> to one file system having an issue, when it is serving (or using) many
>> more file systems. I currently have a box with 5 10Tb file systems on
>> it, and when I mount a 6th file system (2Tb) which I *know* has metadata
>> inconsistencies that fsck can't fix, I don't want it to take down all
>> 50Tb of good solid storage.
>
> This is slightly off-topic, but this is one of the reasons full
> virtualisation is used. If your kernel panics inside a VM, it's
> expendable :)
I'm not sure that makes a lot of sense. Honestly, it's not efficient to
have a single virtual system for each mount point/file system, because
of the overhead. You lose a lot doing that, and only gain the
protection of a trapped panic inside a virtualized host.
At least one major commercial file system vendor that I have a lot of
direct experience with ejects the filesystem from the system if it
incurs a serious issue, and returns i/o errors. Truthfully, it's *FAR*
better than having the kernel panic and taking out the whole system.
This conversation is really a -fs@ discussion, maybe we should move it
there?
Eric
More information about the freebsd-geom
mailing list