1 |
On 09/23/2014 11:28 AM, Mike Gilbert wrote: |
2 |
> On Tue, Sep 23, 2014 at 1:40 PM, walt <w41ter@×××××.com> wrote: |
3 |
>> On 09/23/2014 07:46 AM, Rich Freeman wrote: |
4 |
>>> On Tue, Sep 23, 2014 at 10:27 AM, walt <w41ter@×××××.com> wrote: |
5 |
>>>> |
6 |
>>>> I used systemctl to stop and restart systemd-journald, thinking I might |
7 |
>>>> see some useful error messages. But when systemd-journal started up |
8 |
>>>> again the journal file was back in /var/log/journal where I want it :) |
9 |
>>>> |
10 |
>>>> No idea why rebooting the machine didn't do the same thing. |
11 |
>>>> |
12 |
>>> |
13 |
>>> Are you sure that it is solved, and that the problem won't recur on |
14 |
>>> the next reboot? |
15 |
>> |
16 |
>> <sigh> After a reboot the journal file is back in /run/log/journal. |
17 |
>> |
18 |
>>> If it does, my next question (an educated guess, but a guess) would be |
19 |
>>> whether you're using an initramfs, |
20 |
>> |
21 |
>> No, I never have. |
22 |
>> |
23 |
>>> I'd also look at anything |
24 |
>>> that might be causing issues with /var/log/journal when journald is |
25 |
>>> launched, such as that directory being on an unmounted filesystem and |
26 |
>>> there not being some dependency that causes journald to notice. |
27 |
>> |
28 |
>> This particular machine has only root and swap partitions, so there's |
29 |
>> nothing to remain unmounted during boot. |
30 |
>> |
31 |
>> Having reassured myself with that claim, I now spot this journal message |
32 |
>> (which appears only on the 'broken' machine): |
33 |
>> |
34 |
>> Sep 23 07:40:46 a6 systemd[1]: Found ordering cycle on sysinit.target/start |
35 |
>> Sep 23 07:40:46 a6 systemd[1]: Found dependency on local-fs.target/start |
36 |
>> Sep 23 07:40:46 a6 systemd[1]: Found dependency on lvm.service/start |
37 |
>> Sep 23 07:40:46 a6 systemd[1]: Found dependency on sysinit.target/start |
38 |
>> Sep 23 07:40:46 a6 systemd[1]: Breaking ordering cycle by deleting job local-fs.target/start |
39 |
>> Sep 23 07:40:46 a6 systemd[1]: Job local-fs.target/start deleted to break ordering cycle starting with sysinit.target/start |
40 |
>> |
41 |
>> I don't understand everything about that message, but it seems to imply |
42 |
>> that systemd may think that the local filesystems are not mounted(?) |
43 |
>> |
44 |
>> Could this be causing my journald problem, maybe? |
45 |
> |
46 |
> Where is lvm.service coming from? I suspect it is causing the ordering cycle. |
47 |
|
48 |
Great pickup, thanks Mike. |
49 |
|
50 |
I took lvm.service from either Arch or Fedora way back when gentoo first |
51 |
introduced systemd. For whatever reason my lvm volumes weren't being activated |
52 |
during bootup. The lvm.service file fixed that for a long time and caused no |
53 |
trouble until recently. |
54 |
|
55 |
I just disabled the lvm.service and rebooted: the journal file is back in |
56 |
/var/log/journal where it should be :) |
57 |
|
58 |
Just as important, the lvm filesystem is properly mounted during bootup, so |
59 |
somewhere along the line that problem got fixed independently of systemd. |
60 |
|
61 |
Thanks again Mike, Rich, and Canek. Excellent help and much appreciated. |