From: Adrian Larsen <alarsen@maidenheadbridge.com>
To: wireguard@lists.zx2c4.com
Subject: Re: Continued use of `wg-quick save` and SaveConfig=true?
Date: Tue, 5 Jan 2021 00:16:48 +0000 [thread overview]
Message-ID: <efc2e5c8-a46b-38ad-04c4-1e11c33c021e@maidenheadbridge.com> (raw)
In-Reply-To: <da590a57-f605-07a0-de90-319e7e163786@de-vri.es>
Hi Maarten,
Thanks for your answer..
On 04/01/2021 21:05, Maarten de Vries wrote:
>
> On 04-01-2021 19:41, Adrian Larsen wrote:
>> Hi Jason,
>>
>> 1) From a manual operation point of view, I feel more comfortable if
>> an Operator uses:
>>
>> # wg set wg0 peer ... allowed-ips ...
>> # wg-quick save wg0
>>
>> rather than editing manually the config file.
>>
>> In case the Wire Guard is running multiple peers with production
>> traffic, I think an Operator can do less damage using the commands if
>> something goes wrong.
>
>
> Another solution could be to have a command line tool that modifies
> the configuration file. Kinda like `wg set` except for a config file
> instead of a live interface. Would tooling like that alleviate your
> concerns of an operator messing up the configuration file?
>
Yes, one of them. :-)
The second potential problem is if the change is wrong, it is better not
to permanent save it before to test. This is a common scenario:
1- The operator does a change using "wg set" on "x remote node"
2- The setting is incorrect and the "x remote node" becomes isolated.
The operator cannot access to it any more.
3- Last resource solution: Due to the change was not saved, rebooting "x
remote node" will allow the Operator to have access to it again.
>
>>
>> 2) From automation point of view, still I think that is easy to use
>> the commands (on an script):
>>
>> # wg set wg0 peer ... allowed-ips ...
>> # wg-quick save wg0
>>
>> rather than using "sed" or "awk" to modify the config file.
>
>
> Yeah, sed and awk aren't necessarily the nicest solutions. Although
> they would work fine in practice. But maybe a tool as mentioned above
> could solve this. Even just a script as pretty front-end to the right
> sed/awk commands could be useful here.
>
> For automation, (re-)applying a config file does have an important
> advantage over `wg set ... && wg-quick save ...`: you can be sure that
> all changes are applied, even if the tunnel was temporarily gone for
> some reason. Worst case, the changes are kept in the config file and
> applied when the interface is created again.
>
> If you try to do `wg set ... && wg-quick save ...` while the WireGuard
> interface doesn't exist, the changes are lost. (Or you have to fall
> back to modifying the config file by hand, but we didn't want that.)
>
> This advantage may not matter for your application, I can't really
> know that. I suppose that depends on the kind of changes that are made
> to the configuration at runtime.
>
> -- Maarten
>
>>
>> My 2 cents.
>>
>> Adrian
>>
>> On 04/01/2021 16:16, Maarten de Vries wrote:
>>> On 03-01-2021 20:59, Chris Osicki wrote:
>>>> On Sat, Jan 02, 2021 at 03:37:09PM +0100, Jason A. Donenfeld wrote:
>>>>> Hi,
>>>>>
>>>>> I was thinking recently that most people have switched from a
>>>>> model of
>>>>> updating the runtime configuration and then reading that back into a
>>>>> config file, to editing the config file and then syncing that with
>>>>> the
>>>>> runtime config. In other words, people have moved from doing:
>>>>>
>>>>> # wg set wg0 peer ... allowed-ips ...
>>>>> # wg-quick save wg0
>>>>>
>>>>> To doing:
>>>>>
>>>>> # vim /etc/wireguard/wg0.conf
>>>>> # wg syncconf wg0 <(wg-quick strip wg0)
>>>>>
>>>>> I think this is mostly a positive change too in terms of reliability.
>>>>> Reading back the runtime configuration was always a bit hit or miss,
>>>>> and I suspect that more times than not people have been confused by
>>>>> SaveConfig=true.
>>>>>
>>>>> That raises the question: are there good uses left for
>>>>> SaveConfig=true
>>>>> and `wg-quick save` that warrant keeping the feature around?
>>>>> Temporarily caching a roamed endpoint IP, perhaps, but how helpful is
>>>>> that?
>>>>>
>>>>> I haven't thought too deeply about this in order to be wedded to one
>>>>> outcome over the other yet, but seeing some confusion today,
>>>>> again, in
>>>>> #wireguard over the feature made me wonder.
>>>>>
>>>>> Any opinions on this? Any one on this list actively use this feature
>>>>> and see replacements for it (e.g. syncconf) as clearly inferior?
>>>>>
>>>>> Jason
>>>> Hi Jason
>>>>
>>>> Being an old fashioned Unix admin, ~30 years spent in this job, I
>>>> vote for the traditional way of doing it:
>>>> change the config file and let the application reread it.
>>>> I think the KISS principle is still valid ;-)
>>>
>>> I totally agree. Reloading the config file is much nicer :)
>>>
>>> I also don't need to save roaming endpoints. All WireGuard tunnels I
>>> use have at-least one side with a fixed endpoint. And if that's not
>>> the case I imagine you probably need a more complicated solution
>>> than wg-quick.
>>>
>>>
>>>> Thanks for the excellent software, Jason!
>>>
>>> I also totally agree with this. WireGuard has made my life a lot
>>> easier :)
>>>
>>>
>>> Regards,
>>>
>>> Maarten
>>>
next prev parent reply other threads:[~2021-01-05 0:16 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-02 14:37 Jason A. Donenfeld
2021-01-03 19:59 ` Chris Osicki
2021-01-04 16:16 ` Maarten de Vries
2021-01-04 18:41 ` Adrian Larsen
2021-01-04 21:05 ` Maarten de Vries
2021-01-05 0:16 ` Adrian Larsen [this message]
2021-01-08 10:42 ` Eicke Herbertz
2021-01-05 2:00 ` Michael B. Williams
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=efc2e5c8-a46b-38ad-04c4-1e11c33c021e@maidenheadbridge.com \
--to=alarsen@maidenheadbridge.com \
--cc=wireguard@lists.zx2c4.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).