1 |
Am 15.05.2014 13:50, schrieb covici@××××××××××.com: |
2 |
> Canek Peláez Valdés <caneko@×××××.com> wrote: |
3 |
> |
4 |
>> On Thu, May 15, 2014 at 1:18 AM, Canek Peláez Valdés <caneko@×××××.com> wrote: |
5 |
>>> On Wed, May 14, 2014 at 5:26 PM, <covici@××××××××××.com> wrote: |
6 |
>>> [snip] |
7 |
>>>> |
8 |
>>>> Well, the workaround sort of worked -- it went through the initrd -- I |
9 |
>>>> had debug in the kernel command line, but it did not stop for nothing! |
10 |
>>>> When it went to the real root, however it did not activate any of the |
11 |
>>>> lvm volumes I had except for what I specified in the kernel command |
12 |
>>>> line, causing things not to work well. Also, I noticed that if insisted |
13 |
>>>> on using the predictable network names, even though I have |
14 |
>>>> /etc/udev/rules.d/70-persistent-net.rules and |
15 |
>>>> /etc/udev/rules.d/80-name-slot.rules which work fine in openrc to give |
16 |
>>>> me back my eth* names. So all in all, it was a mess and took me to an |
17 |
>>>> emergency shell and that was the end of that. I did eventually activate |
18 |
>>>> some volumes by lvchange -aay, but obviously that would not work well. |
19 |
>>> |
20 |
>>> OK, I was a little mystified about why dracut-036 worked on my system |
21 |
>>> and 037 didn't. Before I tried any workaround, I wanted to know what |
22 |
>>> changed from the previous version to the current one. |
23 |
>>> |
24 |
>>> So I generated an initramfs with dracut-036-r4 and another one with |
25 |
>>> dracut-037-r1, and I tried to see what changed from one to the other. |
26 |
>>> The answer is surprisingly easy: in /etc/cmdline.d/, the following |
27 |
>>> files where in the 036-r4 version, but not in the 037-r4: |
28 |
>>> |
29 |
>>> 90crypt.conf |
30 |
>>> 90lvm.conf |
31 |
>>> 90mdraid.conf |
32 |
>>> base.conf |
33 |
>>> |
34 |
>>> Te contents of those files are (90crypt.conf is empty): |
35 |
>>> |
36 |
>>> 90lvm.conf |
37 |
>>> rd.lvm.lv=vg/vol1 |
38 |
>>> rd.lvm.lv=vg/vol4 |
39 |
>>> rd.lvm.lv=vg/vol3 |
40 |
>>> |
41 |
>>> 90mdraid.conf |
42 |
>>> rd.md.uuid=f4a59e68:fbe4039f:a39fc86d:e9e91e12 |
43 |
>>> |
44 |
>>> base.conf |
45 |
>>> ro |
46 |
>>> |
47 |
>>> So I just changed my /etc/default/grub file: |
48 |
>>> |
49 |
>>> GRUB_CMDLINE_LINUX="init=/usr/lib/systemd/systemd quiet nosplash |
50 |
>>> rd.lvm.lv=vg/vol1 rd.lvm.lv=vg/vol4 rd.lvm.lv=vg/vol3 |
51 |
>>> rd.md.uuid=f4a59e68:fbe4039f:a39fc86d:e9e91e12" |
52 |
>>> |
53 |
>>> I regenerated my GRUB2 config, and now again my LVM test system works |
54 |
>>> perfectly with the latest dracut version. |
55 |
>> |
56 |
>> I'm an idiot; I didn't saw the documentation about hostonly_cmdline; |
57 |
>> BTW Jc, you used host_cmdline, I think the former is the correct one. |
58 |
>> |
59 |
>> So, to resume: there is no bug, is just that before hostonly_cmdline |
60 |
>> was yes by default, and now is no by default. This change was |
61 |
>> documented, but I failed to notice it (and I think the ebuild in |
62 |
>> Gentoo should print an einfo message). |
63 |
>> |
64 |
>> Anyway, I think that explains all my problems; John, I don't know if |
65 |
>> it will solve yours. Again: did you used "dracut --print-cmdline" to |
66 |
>> get the command line? Also, have you tried to use -H to generate your |
67 |
>> initramfs? And finally, have you tried with --hostonly-cmdline? |
68 |
> |
69 |
> OK, I was looking through the journal output and I think the key to the |
70 |
> lvm's not activating is the following: |
71 |
> 4 12:54:57 ccs systemd[1]: Got notification message for unit |
72 |
> systemd-journald.service |
73 |
> 4 12:54:57 ccs systemd[1]: systemd-journald.service: Got notification |
74 |
> message from PID 1750 (WATCHDOG=1...) |
75 |
> 4 12:54:57 ccs systemd[1]: systemd-journald.service: got WATCHDOG=1 |
76 |
> 4 12:54:57 ccs systemd[1]: Received SIGCHLD from PID 2603 (lvm). |
77 |
> 4 12:54:57 ccs systemd[1]: Child 2602 (lvm) died (code=exited, |
78 |
> status=5/NOTINSSTALLED) |
79 |
> 4 12:54:57 ccs systemd[1]: Child 2603 (lvm) died (code=exited, |
80 |
> status=5/NOTINSSTALLED) |
81 |
> 4 12:54:57 ccs systemd[1]: Child 2610 (lvm) died (code=exited, |
82 |
> status=5/NOTINSSTALLED) |
83 |
> 4 12:54:57 ccs systemd[1]: Job |
84 |
> dev-mapper-linux\x2d\x2dfiles\x2dportage.device/start timed out. |
85 |
> |
86 |
> So what is not installed? |
87 |
|
88 |
My search tells me that this might be a misinterpreted return code. |
89 |
I might repeat myself but the thread gets quite large now: |
90 |
|
91 |
Did you enable lvm2-lvmetad.service or socket (and set use_lvmetad=1 in |
92 |
lvm.conf)? |
93 |
|
94 |
I think you don't have to, I just ask to check. |
95 |
|
96 |
What release of lvm2, btw? |
97 |
|
98 |
> Also, for the first two lines, I get hundreds of thatpair of lines, how |
99 |
> can I prevent such. |
100 |
|
101 |
The PID1 stuff ? |
102 |
|
103 |
> So, between the lvm problem and the udev renaming my eth0 devices these |
104 |
> are the key as to why things are going wrong -- with openrc udev is not |
105 |
> renaming eth0 at all. |
106 |
|
107 |
We'll take care of eth0 as well as soon your box boots correctly ;-) |
108 |
|
109 |
|
110 |
Stefan |