1 |
-----BEGIN PGP SIGNED MESSAGE----- |
2 |
Hash: SHA256 |
3 |
|
4 |
On 07/08/15 12:59 PM, William Hubbs wrote: |
5 |
> On Fri, Aug 07, 2015 at 12:10:56PM -0400, Ian Stakenvicius wrote: |
6 |
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 |
7 |
>> |
8 |
>> On 07/08/15 11:30 AM, William Hubbs wrote: |
9 |
>>> On Thu, Aug 06, 2015 at 08:07:44PM -0400, Ian Stakenvicius |
10 |
>>> wrote: |
11 |
>>>> |
12 |
>>>> Can we get "nofail" immediately in the mount -a variants of |
13 |
>>>> localmount/netmount and expand that in netmount to make the |
14 |
>>>> nfsclient dep be a "use" or a "need" depending on if it's set |
15 |
>>>> or not?? That would imo kill the existing bug that started |
16 |
>>>> all of this too. |
17 |
>>> |
18 |
>>> Sure, I can get the nofail support in pretty quick, and I |
19 |
>>> think that is a feature we should have. |
20 |
>>> |
21 |
>>> Right now, netmount is using the use dependency to make sure |
22 |
>>> network file system utilities are started before us. Because |
23 |
>>> of the all-or-nothing nature of netmount, we can't switch |
24 |
>>> those dependencies to need. It would cause netmount to fail if |
25 |
>>> one of those utilities fails to start. The use dependency is |
26 |
>>> the best one we can use at this time, and a migration path was |
27 |
>>> specifically laid out in the news item. |
28 |
>>> |
29 |
>> |
30 |
>> My thinking here is that, unless network mounts in fstab are |
31 |
>> listed as 'nofail', that netmount failing due to the dependent |
32 |
>> services not being able to start would be a valid case. |
33 |
>> Sysadmins that don't want netmount to fail no matter what would |
34 |
>> be able to use 'nofail' to ensure that happens. |
35 |
>> |
36 |
>> This is of course predicated on (1) it being a good idea, and |
37 |
>> (2) fstabinfo or whatever the check currently is that would add |
38 |
>> nfsclient to depend() could easily swap 'use' for 'need' based on |
39 |
>> the (lack of) existence of the nofail attribute in fstab. |
40 |
> |
41 |
> The issue with using the need dependency is that netmount is not |
42 |
> granular enough. It mounts all types of network file systems, so if |
43 |
> we fail because nfs is a need dependency and doesn't start, no |
44 |
> other types of network file systems that use daemons will be |
45 |
> mounted. That's what I meant by all-or-nothing. |
46 |
> |
47 |
> William |
48 |
> |
49 |
|
50 |
Yes I follow that. My thoughts are #1, netmount (and localmount for |
51 |
that matter) are all-or-nothing things; so if the service fails then I |
52 |
don't think anything can necessarily be assumed as to which mounts |
53 |
succeeded and which don't. #2, if 'nofail' is specified on the nfs |
54 |
filesystem mounts (or there aren't any nfs filesystem mounts) in |
55 |
/etc/fstab then the nfsclient 'need' would be a 'use' at most, which |
56 |
means that there won't be any failures for other net-based filesystems |
57 |
and netmount would succeed (assuming those other network filesystem |
58 |
types don't trigger a failure). |
59 |
|
60 |
I don't know if #2 is enough to keep the userbase happy that expects |
61 |
netmount to succeed in mounting other network filesystems. I think we |
62 |
need to ask. |
63 |
|
64 |
What we would essentially end up with is a situation where 'nofail' on |
65 |
the nfs mount points in fstab provides exactly the same behaviour that |
66 |
netmount has now (assuming other network filesystems succeed), while |
67 |
without 'nofail' netmount will fail if there's an nfs filesystem to |
68 |
mount and nfsclient doesn't start, but will bring in nfsclient even if |
69 |
its not in the runlevel and will succeed when nfsclient does start. |
70 |
|
71 |
|
72 |
|
73 |
-----BEGIN PGP SIGNATURE----- |
74 |
Version: GnuPG v2 |
75 |
|
76 |
iF4EAREIAAYFAlXE7U0ACgkQAJxUfCtlWe3h8wD7BcqKAETwIohU3SsVMWI5keFQ |
77 |
eXYPpYKSZbaygHpzkB4BALFYqAQlK+7lEAV3Yszn77z2/VrEcxbp4gaU1UrhUbS7 |
78 |
=0R1x |
79 |
-----END PGP SIGNATURE----- |