1 |
On Sun, 2012-05-27 at 09:59 +0200, Alan McKinnon wrote: |
2 |
> On Sun, 27 May 2012 09:05:46 +0200 |
3 |
> Jarry <mr.jarry@×××××.com> wrote: |
4 |
> |
5 |
> > I have read through all replies, but I still did not find |
6 |
> > answers to my original questions: |
7 |
> > |
8 |
> > Q1: Can I somehow reduce the size of /run? I know it is tmpfs |
9 |
> > and I know this is upper limit normally never achieved, but |
10 |
> > I want to reduce this upper limit. Is it possible, or is it |
11 |
> > hard-coded to half of physical memory? |
12 |
> |
13 |
> I think this works IIRC: |
14 |
> |
15 |
> List it in /etc/fstab. Max size goes in the options field using the |
16 |
> syntax described in man mount |
17 |
> |
18 |
> |
19 |
> > Q2: Can I turn this "/run in tmpfs" feature off? I do not |
20 |
> > see *any* advantage in vasting memory for /run (although |
21 |
> > I agree there might be some point in moving "run" from |
22 |
> > /var/run to /run). But I see one big problem: |
23 |
> |
24 |
> |
25 |
> If if limit the tmpfs to say 100M or so then this is not a problem at |
26 |
> all |
27 |
> |
28 |
> |
29 |
> > |
30 |
> > If badly written application starts writing some crap in |
31 |
> > /run, it could deadlock my computer quite easily. And before |
32 |
> > you ask, no it is not so easy to do with /run on hard-drive |
33 |
> > because I have plenty of TB there and monitoring software |
34 |
> > running which alerts me as soon as any partition is half |
35 |
> > full. Unfortunatelly this does not work for tmpfs because |
36 |
> > with given read/write speed of ram-disk it would be full |
37 |
> > in a few seconds before I had any chance to act... |
38 |
> > |
39 |
> > Jarry |
40 |
> > |
41 |
> |
42 |
> |
43 |
> |
44 |
|
45 |
all on one line: |
46 |
|
47 |
tmpfs /tmp tmpfs |
48 |
size=2500M,mode=1777,noatime,auto 0 0 |
49 |
|
50 |
|
51 |
4G ram (diskless, atom board) works well |
52 |
3G on an otherwise similar system goes bang when compiling glibc or gcc |
53 |
as portage and portage tmp in /tmp and ram needed for compiling meet in |
54 |
the middle :) |
55 |
|
56 |
Helped by swap on an NDB and mapping some space over NFS when really |
57 |
needed. |
58 |
|
59 |
BillK |