1 |
On Tue, Mar 13, 2012 at 12:49 PM, Florian Philipp <lists@×××××××××××.net> wrote: |
2 |
> Am 13.03.2012 17:26, schrieb Michael Mol: |
3 |
>> On Tue, Mar 13, 2012 at 12:11 PM, Florian Philipp <lists@×××××××××××.net> wrote: |
4 |
>>> Am 13.03.2012 12:55, schrieb Valmor de Almeida: |
5 |
>>>> On 03/11/2012 02:29 PM, Florian Philipp wrote: |
6 |
>>>>> Am 11.03.2012 16:38, schrieb Valmor de Almeida: |
7 |
>>>>>> |
8 |
>>>>>> Hello, |
9 |
>>>>>> |
10 |
>>>>>> I have not looked at encryption before and find myself in a situation |
11 |
>>>>>> that I have to encrypt my hard drive. I keep /, /boot, and swap outside |
12 |
>>>>>> LVM, everything else is under LVM. I think all I need to do is to |
13 |
>>>>>> encrypt /home which is under LVM. I use reiserfs. |
14 |
>>>>>> |
15 |
>>>>>> I would appreciate suggestion and pointers on what it is practical and |
16 |
>>>>>> simple in order to accomplish this task with a minimum of downtime. |
17 |
>>>>>> |
18 |
>>>>>> Thanks, |
19 |
>>>>>> |
20 |
>>>>>> -- |
21 |
>>>>>> Valmor |
22 |
>>>>>> |
23 |
>>>>> |
24 |
>>>>> |
25 |
>>>>> Is it acceptable for you to have a commandline prompt for the password |
26 |
>>>>> when booting? In that case you can use LUKS with the /etc/init.d/dmcrypt |
27 |
>>>> |
28 |
>>>> I think so. |
29 |
>>>> |
30 |
>>>>> init script. /etc/conf.d/dmcrypt should contain some examples. As you |
31 |
>>>>> want to encrypt an LVM volume, the lvm init script needs to be started |
32 |
>>>>> before this. As I see it, there is no strict dependency between those |
33 |
>>>>> two scripts. You can add this by adding this line to /etc/rc.conf: |
34 |
>>>>> rc_dmcrypt_after="lvm" |
35 |
>>>>> |
36 |
>>>>> For creating a LUKS-encrypted volume, look at |
37 |
>>>>> http://en.gentoo-wiki.com/wiki/DM-Crypt |
38 |
>>>> |
39 |
>>>> Currently looking at this. |
40 |
>>>> |
41 |
>>>>> |
42 |
>>>>> You won't need most of what is written there; just section 9, |
43 |
>>>>> "Administering LUKS" and the kernel config in section 2, "Assumptions". |
44 |
>>>>> |
45 |
>>>>> Concerning downtime, I'm not aware of any solution that avoids copying |
46 |
>>>>> the data over to the new volume. If downtime is absolutely critical, ask |
47 |
>>>>> and we can work something out that minimizes the time. |
48 |
>>>>> |
49 |
>>>>> Regards, |
50 |
>>>>> Florian Philipp |
51 |
>>>>> |
52 |
>>>> |
53 |
>>>> Since I am planning to encrypt only home/ under LVM control, what kind |
54 |
>>>> of overhead should I expect? |
55 |
>>>> |
56 |
>>>> Thanks, |
57 |
>>>> |
58 |
>>> |
59 |
>>> What do you mean with overhead? CPU utilization? In that case the |
60 |
>>> overhead is minimal, especially when you run a 64-bit kernel with the |
61 |
>>> optimized AES kernel module. |
62 |
>> |
63 |
>> Rough guess: Latency. With encryption, you can't DMA disk data |
64 |
>> directly into a process's address space, because you need the decrypt |
65 |
>> hop. |
66 |
>> |
67 |
> |
68 |
> Good call. Wouldn't have thought of that. |
69 |
> |
70 |
>> Try running bonnie++ on encrypted vs non-encrypted volumes. (Or not; I |
71 |
>> doubt you have the time and materials to do a good, meaningful set of |
72 |
>> time trials) |
73 |
>> |
74 |
> |
75 |
> Yeah, that sounds like something for which you need a very dull winter |
76 |
> day. Besides, I've already lost a poorly cooled HDD on a benchmark. |
77 |
|
78 |
Sounds like something we can do at my LUG at one of our weekly |
79 |
socials. The part I don't know is how to set this kind of thing up and |
80 |
how to tune it; I don't want it to be like Microsoft's comparison of |
81 |
SQL Server against MySQL from a decade or so ago, where they didn't |
82 |
tune MySQL for their bench workload. |
83 |
|
84 |
-- |
85 |
:wq |