1 |
On Sunday, May 04, 2014 09:03:08 PM Stefan G. Weichinger wrote: |
2 |
> Am 04.05.2014 20:40, schrieb J. Roeleveld: |
3 |
> > On Sunday, May 04, 2014 03:07:28 PM Stefan G. Weichinger wrote: |
4 |
> >> Oh, yes, I like ZFS and its features and used it in some cases already. |
5 |
> >> But I didn't yet take the step to set up ZFS-root on my work machines. |
6 |
> > |
7 |
> > I haven't yet, but it's on the list of items to look into at some point. |
8 |
> > I'd like to know if, with ZFS, it is possible to create block-devices like |
9 |
> > LVs which I can then attach to VMs. Or if I have to use files instead. |
10 |
> I think you would have to use files on top of ZFS ... but I am not |
11 |
> up-to-date in that area. |
12 |
> |
13 |
> People run KVM-hosts with storage on ZFS ... nice with the snapshots etc ... |
14 |
|
15 |
Does KVM support running snapshots? |
16 |
Eg. also storing the memory and registers? |
17 |
I have not found any indication that KVM supports that. |
18 |
|
19 |
Without that, KVM is useless to me. |
20 |
|
21 |
|
22 |
> > That's a hardware raid device with 4 * 3TB disks with raid-6. |
23 |
> > I don't like the fact that a 2nd disk-failure can kill a raid-10 when both |
24 |
> > disks are in the same mirror-set. |
25 |
> > |
26 |
> > gdisk automatically aligns on 2048 sector boundaries, that is more then |
27 |
> > enough for the 4k-sectors and the block/stripe sizes employed by the |
28 |
> > raid-controller. |
29 |
> Yes, this is for UEFI booting, thanks. |
30 |
> |
31 |
> I tried to migrate to GPT/BIOS-booting today but failed. It seems my |
32 |
> mainboard/BIOS has problems detecting that ... I vaguely remember this |
33 |
> from trying it back then. |
34 |
|
35 |
I thought the MBR info that's possible with GPT would make it possible with |
36 |
any BIOS? |
37 |
Provided the /boot partition is early enough on the disk. |
38 |
|
39 |
> So I am back on a freshly partitioned and formatted SSD with plain old |
40 |
> MBR now. |
41 |
> |
42 |
> I also wanted to partition the SSD according to the Erase Block Size of |
43 |
> 6144 kB by this way ... dunno if this is still needed or has any real |
44 |
> speed benefits. |
45 |
|
46 |
Again, I would expect current tools should do that automagically? |
47 |
|
48 |
> Maybe I take another approach to migrate to UEFI/GPT in the next days, |
49 |
> now that I have my rsynced filesystems at hand (I got rid of more LVs |
50 |
> and stuff so it gets pretty slim now). |
51 |
|
52 |
Less LVs is simpler. More is more flexible. |
53 |
|
54 |
> > UUIDs, I believe, do work natively. And those are stored inside the |
55 |
> > partition itself. Which means they should also work. But are not as easy |
56 |
> > to locate. (eg. you don't specify them yourself) |
57 |
> |
58 |
> Yep. LABELs are human readable ... big advantage. |
59 |
|
60 |
I would like LABELs to be supported directly by the kernel. |
61 |
|
62 |
> > See the partitioning on my server above. |
63 |
> > It boots using BIOS as I haven't been able to boot Xen using UEFI yet. |
64 |
> |
65 |
> So then the EFI partition is useless ... |
66 |
|
67 |
True, but I leave it there as repartitioning requires extended downtime. And I |
68 |
do occasionally test new versions to see if I can get it to work. |
69 |
|
70 |
> > Support should be there now, but not been able to test that yet. |
71 |
> > GPT is supported by grub-1 (and grub2) and with the MBR-support inside |
72 |
> > GPT, |
73 |
> > booting works from BIOS/MBR. |
74 |
> |
75 |
> ... if your BIOS isn't crappy ;-) |
76 |
|
77 |
Try updating? :) |
78 |
|
79 |
Seriously, do you have the following: |
80 |
|
81 |
*** |
82 |
# man gdisk |
83 |
artemis ~ # gdisk -l /dev/sda |
84 |
GPT fdisk (gdisk) version 0.8.8 |
85 |
|
86 |
Partition table scan: |
87 |
MBR: protective |
88 |
|
89 |
*** |
90 |
|
91 |
That last line of what I copied is the bit that should make it possible. |
92 |
|
93 |
Also, the /boot partition needs to have the mbr-boot flag (or whatever it's |
94 |
called) enabled. |
95 |
|
96 |
-- |
97 |
Joost |