1 |
Am 11.06.2014 22:17, schrieb thegeezer: |
2 |
> On 06/11/2014 07:57 PM, Stefan G. Weichinger wrote: |
3 |
>> looks promising: |
4 |
>> |
5 |
> |
6 |
> awesome. i did have a look through the diff, there are lots of scsi |
7 |
> drivers selected, storage (block) cgroups but i think the crucial factor |
8 |
> was the HZ was set at 100 previously and 1000 now. i guess it has |
9 |
> helped kernel-io though maybe a kernel hacker in here might give a more |
10 |
> authoritative answer |
11 |
|
12 |
|
13 |
The help suggests to choose 100 for servers ... the 1000 comes from the |
14 |
sysresccd-setup, yes. |
15 |
|
16 |
I wonder if chosing the "Processor Family" also had an influence. |
17 |
|
18 |
And I even wonder more if I have some bad choices in my desktop's kernel |
19 |
as well ;-) it's a grown setup over years ... |
20 |
|
21 |
For now it looks good ... I will configure the current kernel as a |
22 |
fallback kernel and maybe play with some options. |
23 |
|
24 |
|
25 |
>> One big fat hw-RAID10 might be better? |
26 |
>> But losing the wrong 2 drives makes it crash again ... afaik. |
27 |
> yeah you could argue with raid6 you can _only_ lose two disks, whereas |
28 |
> if you lose the right disks with raid01 you can lose 3 and still rebuild. |
29 |
> raid 0+1 (as opposed to raid10, slightly different) gives you great |
30 |
> speed and at least one drive you can lose. |
31 |
> however, you are not protected by silent bit corruption but then you are |
32 |
> using btrfs elsewhere. |
33 |
|
34 |
... for the OS, yes ... and maybe for the target of virt-backup. |
35 |
|
36 |
> myself i would use lvm to partition and then at least you can move |
37 |
> things around later; btrfs lets you do the same afaiu |
38 |
> _always_ have your hotspare in the system, then it takes less time to |
39 |
> come back up to 100% |
40 |
> nothing is quite as scary as having a system waiting on the post and a |
41 |
> screwdriver before rebuild can even start |
42 |
|
43 |
good suggestion, sure. |
44 |
|
45 |
That would mean rebuilding the arrays to a RAID6 over 4 or 5 disks and |
46 |
keeping one aside. |
47 |
|
48 |
|
49 |
>> time for a break here. |
50 |
> i'd strongly recommend such monitoring software as munin to have running |
51 |
> -- this way you can watch trends like io times increasing over time and |
52 |
> act on them before things start feeling sluggish |
53 |
|
54 |
I will take a look into it and check how much time I need to learn and |
55 |
set up. |
56 |
|
57 |
> well earned break :) |
58 |
|
59 |
;-) |
60 |
|
61 |
S |