| 1 |
On Sunday 29 August 2010 03:24:42 Nikos Chantziaras wrote: |
| 2 |
> On 08/28/2010 10:42 PM, Dale wrote: |
| 3 |
> > Alex Schuster wrote: |
| 4 |
> >> Dale writes: |
| 5 |
> >>> It would be nice if something like *fdisk could edit the labels tho. |
| 6 |
> >>> It would be so much easier. I didn't see anything in the man pages |
| 7 |
> >>> tho. |
| 8 |
> >> |
| 9 |
> >> I'd like this, too. cfdisk displays them, but is not abel to edit. |
| 10 |
> >> |
| 11 |
> >>> I looked into LVM a good while ago. It's just to much for me to keep |
| 12 |
> >>> up with since I just have a desktop system here. It has its good |
| 13 |
> >>> points but just way overkill for what I have here. |
| 14 |
> >> |
| 15 |
> >> It's not that complicated. In a nutshell: |
| 16 |
> >> |
| 17 |
> >> Choose the partitions you want to use for LVM, and prepare them to be |
| 18 |
> >> physical volumes: |
| 19 |
> >> pvcreate /dev/sda[678] |
| 20 |
> >> |
| 21 |
> >> Create a volume group out of these partitions: |
| 22 |
> >> vgcreate myvg /dev/sda[678] |
| 23 |
> >> |
| 24 |
> >> Create logical volumes in this volume group: |
| 25 |
> >> lvcreate -L 5G -n lvm1 myvg |
| 26 |
> >> lvcreate -L 2G -n lvm2 myvg |
| 27 |
> >> |
| 28 |
> >> Use these logical volumes just as disk partitions: |
| 29 |
> >> |
| 30 |
> >> mke2fs -j -L fs_on_lvm /dev/myvg/lvm1 |
| 31 |
> >> mount /dev/myg/lvm1 /mnt/fs_on_lvm |
| 32 |
> >> |
| 33 |
> >> The file system is too small? Just extend its size by 1G, without |
| 34 |
> >> unmouning: |
| 35 |
> >> |
| 36 |
> >> lvresize -L +1G /dev/myvg/lvm1 |
| 37 |
> >> |
| 38 |
> >> The volume groups is getting full, no space to add LVMs? Add other |
| 39 |
> >> partitions. If you like, even from a 2nd drive: |
| 40 |
> >> |
| 41 |
> >> pvcreate /dev/sdb5 |
| 42 |
> >> vgextend myvg /dev/sdb5 |
| 43 |
> >> |
| 44 |
> >> So, it's of course more complicated than just firing up cfdisk, create |
| 45 |
> >> partitions and file systems on them, but you have much more flexibility. |
| 46 |
> >> Once you have LVM, you do not have to care what the actual device |
| 47 |
> >> names of |
| 48 |
> >> your drives are. If sda becomes sdb and vice versa, no problem, and |
| 49 |
> >> nothing to worry about. LVM does not use the device name, it scans each |
| 50 |
> >> partition and uses the LVM UUIDs on them to identify what is what. |
| 51 |
> >> |
| 52 |
> >> Wonko |
| 53 |
> > |
| 54 |
> > Since I finally got this thing settled with partition sizes, that's |
| 55 |
> > pretty complicated. I have root, /boot, /home, portage and a data |
| 56 |
> > partition for misc. junk. I doubt it will change any in the near future. |
| 57 |
> > |
| 58 |
> > I did read up on it one time a while back. It's neat when you have to |
| 59 |
> > add drives and resize things but still a bit much for a little desktop. |
| 60 |
> |
| 61 |
> I'd stay away from LVM. I started using it on a Debian Lenny machine |
| 62 |
> and performance went down the drain. For example, deleting a 3GB file |
| 63 |
> was almost instant and now it takes like 15 seconds. It's almost as if |
| 64 |
> with LVM, deleting a file means writing 0 all over the 3GB first :-/ |
| 65 |
|
| 66 |
That sounds like a different issue. |
| 67 |
I haven't noticed any major performance issues myself. |
| 68 |
|
| 69 |
But to test quickly: |
| 70 |
LVM: |
| 71 |
# ~/speedtest $ time dd if=/dev/zero of=3gigfile bs=1024 count=3000000 |
| 72 |
3000000+0 records in |
| 73 |
3000000+0 records out |
| 74 |
3072000000 bytes (3.1 GB) copied, 33.3029 s, 92.2 MB/s |
| 75 |
|
| 76 |
real 0m33.305s |
| 77 |
user 0m0.440s |
| 78 |
sys 0m16.370s |
| 79 |
# ~/speedtest $ time rm 3gigfile |
| 80 |
|
| 81 |
real 0m3.827s |
| 82 |
user 0m0.000s |
| 83 |
sys 0m1.131s |
| 84 |
|
| 85 |
# hdparm -Tt /dev/sda |
| 86 |
|
| 87 |
/dev/sda: |
| 88 |
Timing cached reads: 4758 MB in 2.00 seconds = 2379.87 MB/sec |
| 89 |
Timing buffered disk reads: 274 MB in 3.02 seconds = 90.84 MB/sec |
| 90 |
************** |
| 91 |
Non-LVM: |
| 92 |
# /data/speedtest $ time dd if=/dev/zero of=3gigfile bs=1024 count=3000000 |
| 93 |
3000000+0 records in |
| 94 |
3000000+0 records out |
| 95 |
3072000000 bytes (3.1 GB) copied, 38.2821 s, 80.2 MB/s |
| 96 |
|
| 97 |
real 0m38.284s |
| 98 |
user 0m0.397s |
| 99 |
sys 0m9.490s |
| 100 |
# /data/speedtest $ time rm 3gigfile |
| 101 |
|
| 102 |
real 0m0.721s |
| 103 |
user 0m0.000s |
| 104 |
sys 0m0.720s |
| 105 |
|
| 106 |
# hdparm -Tt /dev/sdb |
| 107 |
|
| 108 |
/dev/sdb: |
| 109 |
Timing cached reads: 3396 MB in 2.00 seconds = 1698.30 MB/sec |
| 110 |
Timing buffered disk reads: 252 MB in 3.00 seconds = 83.94 MB/sec |
| 111 |
|
| 112 |
|
| 113 |
Both filesystems are ext3 |
| 114 |
|
| 115 |
Based on this, it takes about 3 seconds more. That is something I can easily |
| 116 |
live with. |
| 117 |
But instantaneous to 15 seconds, I think there might be some other factors |
| 118 |
there? |
| 119 |
|
| 120 |
-- |
| 121 |
Joost |
Archives