1 |
On Wed, 23 May 2012 22:54:23 +0100 |
2 |
Markos Chandras <hwoarang@g.o> wrote: |
3 |
|
4 |
> On 05/23/2012 10:47 PM, Alan McKinnon wrote: |
5 |
> > On Wed, 23 May 2012 22:25:37 +0100 |
6 |
> > Markos Chandras <hwoarang@g.o> wrote: |
7 |
> > |
8 |
> >> On 05/23/2012 05:24 PM, Tanstaafl wrote: |
9 |
> >>> On 2012-05-21 5:00 PM, Markos Chandras <hwoarang@g.o> |
10 |
> >>> wrote: |
11 |
> >>>> On 05/21/2012 03:27 PM, Michael Hampicke wrote: |
12 |
> >>>>>> I updated udev from 171-r5 to 171-r6 and now i get several |
13 |
> >>>>>> udevd boot message as : udevd[1389]: can not find |
14 |
> >>>>>> '/lib/udev/rules.d/90-network.rules': No such file or directory |
15 |
> >>>>>> udevd[1389]: can not find '/lib/udev/rules.d/95-keymap.rules': |
16 |
> >>>>>> No such file or directory ...................... and so on. |
17 |
> >>>>>> |
18 |
> >>>>>> /lib is a symlink pointing to /lib64. /lib64/udev/rules.d is ok |
19 |
> >>>>>> with all the rules that udevd does not find at boot. |
20 |
> >>>>> |
21 |
> >>>>> No I would guess it was because of the upgrade of |
22 |
> >>>>> sys-apps/baselayout to 2.1-r1. Things got crazy here with that |
23 |
> >>>>> upgrade. I had to re-merge every package with files under /lib/ |
24 |
> >>>>> In your case re-merging udev should to the trick. |
25 |
> >>> |
26 |
> >>>> The package clearly informed you that you need to reboot for |
27 |
> >>>> things to work properly |
28 |
> >>>> |
29 |
> >>>> "You should reboot the system now to get /run mounted with |
30 |
> >>>> tmpfs!" |
31 |
> >>>> |
32 |
> >>>> Have a look on pkg_postinst() function in that ebuild. You chose |
33 |
> >>>> to ignore it and this is why you had these problems after the |
34 |
> >>>> update. |
35 |
> >>> |
36 |
> >>> <pet-peeve> |
37 |
> >>> I asked about this a while back but never got a decent answer... |
38 |
> >>> |
39 |
> >>> *Especially* for servers, there really, REALLY needs to be a way |
40 |
> >>> to see this kind of warning BEFORE updating... ie, the warning |
41 |
> >>> should be printed to the screen during an 'emerge -pvuDN world' or |
42 |
> >>> something, so I know that a reboot will be required for this |
43 |
> >>> update. </pet-peeve> |
44 |
> >>> |
45 |
> >> This kind of messages are also printed at the end of -uDNav world |
46 |
> >> so if you scroll your screen up you can see all the warning/log |
47 |
> >> messages from every package that you have updated. Also, these |
48 |
> >> kind of messages are logged in /var/log/portage/ |
49 |
> > |
50 |
> > You are missing the point. |
51 |
> > |
52 |
> > Tanstaafl wants to know if a reboot *will* be required *before* he |
53 |
> > does the update. What you are describing tells him that after the |
54 |
> > update completes when it is already too late. |
55 |
> > |
56 |
> > I face the same issue at work. We have a change policy requiring 14 |
57 |
> > days advance notice of any change affecting service. If I do a |
58 |
> > routine world update then have to log an emergency change for an |
59 |
> > unexpected reboot, the change manager will have my nuts for |
60 |
> > breakfast. |
61 |
> > |
62 |
> > If it happens more than once, I'd be having a really unusual |
63 |
> > conversation with the CTO which probably ends with him standing |
64 |
> > behind me watching while I migrate every single box that isn't |
65 |
> > RHEL6 (all 200 of them) over to RHEL6 where I *do* have exact |
66 |
> > knowledge in advance of the impact of a change. |
67 |
> > |
68 |
> > |
69 |
> > |
70 |
> Did either of you ever open a bug about this or even discuss it in the |
71 |
> gentoo-dev mailing list? What you say sounds like a valid concern to |
72 |
> me but unless you express your needs to maintainers, nothing is ever |
73 |
> going to happen. However, in this particular case, yes a news item |
74 |
> would be the ideal solution. |
75 |
|
76 |
|
77 |
I haven't opened a bug myself, mostly because I've never been bitten |
78 |
by this. My Gentoo servers run stable so I've always known from this |
79 |
list and other places when something requiring a reboot is coming down |
80 |
the line. |
81 |
|
82 |
I agree, a news item is the perfect solution. Having portage do it will |
83 |
be highly cumbersome, it will require some kind of new magic flag in |
84 |
ebuilds that portage must parse. All that work for something that |
85 |
doesn't happen often? Nah, it'll never fly. |
86 |
|
87 |
|
88 |
|
89 |
-- |
90 |
Alan McKinnnon |
91 |
alan.mckinnon@×××××.com |