[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[linux-lvm] pvdisplay -m



is "pvdisplay -m" still unimplemented?

After I turned to a hardware raid5 system my PVs become
quite HUGE (about 2-4 TB  each!). This huge PVs easily
get fragmented. If I want to pvmove some LV it often does not
work by default. The huge PV has space to hold the additional
LV, but it is not contiguous, and "--alloc anywhere" won't
split a contiguous chunk to fit into smaller gaps. (and the
algorithm to do so is not trivial, even if it sounds easy).

So I frequently end up with a desk calculator to analyze my
/etc/lvm/backup/ metadata to extract a list of occupied and
free ranges on my PV to express pvmove commands with explicit
PE ranges.

To make this easier, I would propose the following extensions:

Some command/option to print out the layout of a PV including
the free gaps.

Extend the --units option to display LE counts. (one could use "e",
but I don't know, how long it takes until we need to express sizes
as exabytes :-) 

Express LE ranges by a "+" like /dev/sda:1000+500 as a shortcut
for /dev/sda:1000-1499.

btw: is there a more detailed description about how pvmove exactly
behaves when giving multiple source and/or target ranges?
--> can I explicitly express the splitting of source chunks into
target fragments by a single pvmove command?

If someone agrees that this is useful, I may start trying to
implement this and propose some patches.

Dieter.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]