1 |
On 12/06/2018 02:27 AM, Dale wrote: |
2 |
> From what I've read, I can use pvmove and pvremove to replace that drive. |
3 |
> Just tell pv to move the data and when done, remove the old drive. After |
4 |
> that, the new 6TB drive will be used in that PV and the 3TB drive can |
5 |
> be used for something else. Is it really that easy or is there more to |
6 |
> it than that? Pardon me but that doesn't sound complicated enough to me. |
7 |
|
8 |
I've migrated multiple hundreds of TB of data this way. |
9 |
|
10 |
In short: |
11 |
|
12 |
1) Partition the new drive(s) as desired. |
13 |
2) pvcreate /dev/$newPv |
14 |
3) vgextend $vgName /dev/$newPv |
15 |
4) pvmove /dev/$oldPv /dev/$newPv |
16 |
5) vgreduce $vgName /dev/$oldPv |
17 |
6) pvremove /dev/$oldPv |
18 |
|
19 |
This does work well, even if the LV(s) are in use / file system(s) are |
20 |
mounted. |
21 |
|
22 |
I have occasionally had issues where the system seems to not respond, |
23 |
despite the fact that it is doing what it's supposed to. I wonder if |
24 |
it's related to the memory leak that J. Roeleveld was talking about. |
25 |
|
26 |
Note: I /do/ *STRONGLY* recommend that you do partition the new drive |
27 |
and /not/ pvcreate the entire drive. — Many of the data recovery tools |
28 |
/expect/ there to be a partition table. Those that don't care are happy |
29 |
to work with a partition table. I've seen others be in a very |
30 |
uncomfortable situation when they /didn't/ use a partition table. |
31 |
Simple easy thing to avoid painting yourself into a corner. |
32 |
|
33 |
|
34 |
|
35 |
-- |
36 |
Grant. . . . |
37 |
unix || die |