On Fri, Sep 19, 2008 at 15:39, jonathan sai <Jonathan.Sai at sun.com>
wrote:> Hi All,
>
> Need to trouble you to open up the pdf. Customer has included int he doc
> configuration that they are having and at the end their query. Appreciate
> any help to respond to the customer.
First, it''s not recommended to run non-redundant pools (as all of the
pools in the PDF are). This prevents ZFS from recovering from errors:
it can still detect incorrect checksums, but when it finds one it has
no way to get good data to replace the missing or damaged blocks.
Running mirrored or raidz vdevs will allow ZFS to fix problems of this
sort that it finds.
As for the questions:
1) How do I know how much space I am adding?
Given the number of disks N in a vdev, of minimum capacity C:
Mirroring gives additional capacity C
Raidz gives additional capacity (N-1)*C
Raidz2 gives additional capacity (N-2)*C
* How do I check much space does one disk device have?
Check ''format'' output; many devices show how big they are in
the
summary-of-disks screen.
* How do I know if the disk device is in use before I use it to add
into a existing
pool?
Trying "zpool add -n mypool cntndn" will tell you definitively whether
the disk is "in use" from ZFS'' point of view.
2) Is it safer to do a scrub when there are no users using it? i.e:
when it is offline?
It''s not safer in the loss-of-data sense; ZFS will not destroy any
data that isn''t already broken by running a scrub. But scrubbing does
a whole lot of I/O, and this can hurt performance; at work several of
our disks (which serve user home directories) are attached via a
single U320 SCSI channel (which may or may not be the problem), and
running a scrub during business hours brings complaints from users.
Will