[gentoo-user] Question about bird init script

[gentoo-user] Question about bird init script

[Alarig Le Lay]

Hi again,

On another box, I’m trying to update to bird 2.0.4. I just copied the
ebuild and changed the version. If you want to have a look, it’s here:
https://git.grifon.fr/alarig/SwordArMor-gentoo-overlay/src/branch/master/net-misc/bird

It compiles well, but I have a weird behaviour on the service if I want
to stop it.
judicael-ovpn2 ~ # rc-service bird stop
 * Stopping BIRD ...
 * start-stop-daemon: 1 process refused to stop
 * Failed to stop BIRD                                                    [ !! ]
 * ERROR: bird failed to stop

But the process is actually killed and the service marked as crashed.
judicael-ovpn2 ~ # ps aux | grep bird
root     16679  0.0  0.1  10860  2152 pts/1    R+   10:48   0:00 grep --colour=auto bird
judicael-ovpn2 ~ # rc-service bird status
 * status: crashed


Otherwise, if I restart it, it works well.
judicael-ovpn2 ~ # rc-service bird zap
 * Manually resetting bird to stopped state
judicael-ovpn2 ~ # rc-service bird restart
 * Starting BIRD ...                                                      [ ok ]
judicael-ovpn2 ~ # ps aux | grep bird
root     16951 50.1  1.7  40304 35128 ?        Rs   10:49   0:03 /usr/sbin/bird -c /etc/bird.conf -s /var/run/bird.ctl
root     17063  0.0  0.1  10860  2168 pts/1    S+   10:49   0:00 grep --colour=auto bird
judicael-ovpn2 ~ # birdc 'sh pr'
BIRD 2.0.4 ready.
Name       Proto      Table      State  Since         Info
device1    Device     ---        up     10:49:03.725
direct1    Direct     ---        up     10:49:03.725
kernel_ipv4 Kernel     master4    up     10:49:03.725
kernel_ipv6 Kernel     master6    up     10:49:03.725
ibgp_nominoe_ipv4 BGP        ---        up     10:49:05.942  Established
ibgp_nominoe_ipv6 BGP        ---        up     10:49:05.942  Established
ibgp_budic_ipv4 BGP        ---        up     10:49:07.189  Established
ibgp_budic_ipv6 BGP        ---        up     10:49:07.656  Established
ospf_ipv4  OSPF       master4    up     10:49:03.725  Running
ospf_ipv6  OSPF       master6    up     10:49:03.725  Running


I looked at the init script (in the files dir of the ebuild), but I
didn’t see what could fail.

Thanks,
-- 
Alarig

Re: [gentoo-user] Question about bird init script

[Michael Orlitzky]

On 3/4/19 5:33 AM, Alarig Le Lay wrote:
> Hi again,
> 
> On another box, I’m trying to update to bird 2.0.4. I just copied the
> ebuild and changed the version. If you want to have a look, it’s here:
> https://git.grifon.fr/alarig/SwordArMor-gentoo-overlay/src/branch/master/net-misc/bird
> 
> It compiles well, but I have a weird behaviour on the service if I want
> to stop it.

I know this isn't the answer you want, but before you waste any time
debugging, you should rewrite that ugly-ass service script to meet
modern standards:

  https://github.com/OpenRC/openrc/blob/master/service-script-guide.md

I'm pretty sure you can delete 90% of it, and make it work better in the
process. Then, if start/stop/reload still don't work, it will be
worthwhile to figure out why.

Re: [gentoo-user] Question about bird init script

[Alarig Le Lay]

Okay, I tried to write another script:
https://git.grifon.fr/alarig/SwordArMor-gentoo-overlay/src/branch/master/net-misc/bird/files/initd-bird-2

I deleted things, but added others. I’m not sure if I respect all the
openrc standards as I’m not very comfortable with it.

I added a check_run() function because I want to use the bird config
parser if it’s running.

I added my own stop() function because bird has to wait for all the
sessions to be closed before killing the process.

I don’t know if I should make the configuration file and the sock file
configurable with /etc/conf.d/bird

I’m open to your comments and improvements :)


My script behave like this:

judicael-ovpn2 ~ # ps aux | grep bird
root      3928  0.0  0.3  25752  7944 pts/1    S+   Mar04   0:04 vim /etc/init.d/bird
root      5250  0.0  0.1  10860  2052 pts/3    S+   11:31   0:00 grep --colour=auto bird
judicael-ovpn2 ~ # rc-service bird start
 * Starting bird ...                                                      [ ok ]
judicael-ovpn2 ~ # ps aux | grep bird
root      3928  0.0  0.3  25752  7944 pts/1    S+   Mar04   0:04 vim /etc/init.d/bird
root      5383 62.8  1.8  42184 37080 ?        Rs   11:31   0:03 /usr/sbin/bird -c /etc/bird.conf -s /var/run/bird.ctl -P /var/run/bird.pid
root      5400  0.0  0.1  10860  2236 pts/3    S+   11:31   0:00 grep --colour=auto bird
judicael-ovpn2 ~ # birdc 'sh pr'
BIRD 2.0.4 ready.
Name       Proto      Table      State  Since         Info
device1    Device     ---        up     11:31:26.782
direct1    Direct     ---        up     11:31:26.782
kernel_ipv4 Kernel     master4    up     11:31:26.782
kernel_ipv6 Kernel     master6    up     11:31:26.782
ibgp_nominoe_ipv4 BGP        ---        up     11:31:28.777  Established
ibgp_nominoe_ipv6 BGP        ---        up     11:31:28.205  Established
ibgp_budic_ipv4 BGP        ---        up     11:31:30.703  Established
ibgp_budic_ipv6 BGP        ---        up     11:31:28.614  Established
ospf_ipv4  OSPF       master4    up     11:31:26.782  Running
ospf_ipv6  OSPF       master6    up     11:31:26.782  Running
judicael-ovpn2 ~ # rc-service bird reload
 * Reloading BIRD ...                                                     [ ok ]
judicael-ovpn2 ~ # rc-service bird restart
 * Stopping BIRD ...
 * Starting bird ...                                                      [ ok ]
judicael-ovpn2 ~ # rc-service bird status
 * status: started


So basically it works, and the error check as well:
judicael-ovpn2 ~ # echo 'ezf' >> /etc/bird.conf
judicael-ovpn2 ~ # rc-service bird reload
 * /etc/bird.conf:121:1 syntax error, unexpected SYM
judicael-ovpn2 ~ # rc-service bird restart
 * Caching service dependencies ...                                       [ ok ]
 * /etc/bird.conf:121:1 syntax error, unexpected SYM
 * ERROR: bird failed to stop
judicael-ovpn2 ~ # sed -i '$d' /etc/bird.conf
judicael-ovpn2 ~ # rc-service bird reload
 * Reloading BIRD ...                                                     [ ok ]

-- 
Alarig

Re: [gentoo-user] Question about bird init script

[Michael Orlitzky]

On 3/5/19 6:03 AM, Alarig Le Lay wrote:
> Okay, I tried to write another script:
> https://git.grifon.fr/alarig/SwordArMor-gentoo-overlay/src/branch/master/net-misc/bird/files/initd-bird-2
> 
> I deleted things, but added others. I’m not sure if I respect all the
> openrc standards as I’m not very comfortable with it.

That's OK, it already looks a lot better. And it works, which is nice =)


> 
> I added a check_run() function because I want to use the bird config
> parser if it’s running.

You probably want to specify the config file in check_run(). You might
refactor the command_args as follows,

  client_args="-c ${CONF_FILE} -s ${SOCK}"
  command_args="${client_args} -P ${pidfile}"

and then in check_run, you can use "birdc ${client_args}..."


> 
> I added my own stop() function because bird has to wait for all the
> sessions to be closed before killing the process.

There's a "retry" variable that you can set at the top-level:

  retry          Retry schedule to use when stopping the daemon. It can
                 either be a timeout in seconds or multiple signal/time‐
                 out pairs (like SIGTERM/5).

  (from "man openrc-run")

The default stop() function will use that, so you shouldn't need to
write your own stop() if you set "retry=15".


> 
> I’m open to your comments and improvements :)
> 

The checkconfig() function isn't used any more, so it can be deleted. So
long as bird crashes with an error like "no config file!" when the file
is missing, I don't think you need to check for it yourself.

These days /var/run is simply a symlink to /run, so those two paths can
be shortened.

Once you've used it for a while and are confident that everything works,
please file a bug to include the improved script in the tree!

Re: [gentoo-user] Question about bird init script

[Alarig Le Lay]

Thanks for you help!

I just did one think differently, I explain it inline.
I will run bird 2.0.4 on my lab for the rest of the week, then I will
upgrade the production network gradually. I will propose my script to
the tree if it all succeeds.

On mar.  5 mars 10:01:08 2019, Michael Orlitzky wrote:
> On 3/5/19 6:03 AM, Alarig Le Lay wrote:
> > Okay, I tried to write another script:
> > https://git.grifon.fr/alarig/SwordArMor-gentoo-overlay/src/branch/master/net-misc/bird/files/initd-bird-2
> > 
> > I deleted things, but added others. I’m not sure if I respect all the
> > openrc standards as I’m not very comfortable with it.
> 
> That's OK, it already looks a lot better. And it works, which is nice =)

Hehe :)

> You probably want to specify the config file in check_run(). You might
> refactor the command_args as follows,
> 
>   client_args="-c ${CONF_FILE} -s ${SOCK}"
>   command_args="${client_args} -P ${pidfile}"
> 
> and then in check_run, you can use "birdc ${client_args}..."

The client doesn’t care about the configuration file, only about the
socket. However, I factored the socket and then added -r to restrict
read-only commands, in order to avoid protocol disabling during the
check.
“Option -r can be used to enable a restricted mode of BIRD client, which
allows just read-only commands (show ...).”
 -- https://bird.network.cz/?get_doc&v=20&f=bird-4.html

  client_args="-s ${SOCK}"
  command_args="${client_args} -c ${CONF_FILE} -P ${pidfile}"
  client_args="${client_args} -r"

I also factored my `birdc config check` command in check_run().

-- 
Alarig

Re: [gentoo-user] Question about bird init script

[Michael Orlitzky]

On 3/5/19 11:24 AM, Alarig Le Lay wrote:
> 
> The client doesn’t care about the configuration file, only about the
> socket. 

Oh, sorry, I was reading the wrong part of the man page.

It looks like you *can* specify which config file to check, but in a
different way:

  configure check ["config file"]

    Read and parse given config file, but do not use it. useful for
    checking syntactic and some semantic validity of an config file.

So you'd want something like "configure check ${CONF_FILE}" I guess.
That way if you ever have two instances of bird running, reloading the
second instance won't check the config file for the first instance.

And "-r" was a good idea.

Re: [gentoo-user] Question about bird init script

[Alarig Le Lay]

On mar.  5 mars 14:45:46 2019, Michael Orlitzky wrote:
> Oh, sorry, I was reading the wrong part of the man page.
> 
> It looks like you *can* specify which config file to check, but in a
> different way:
> 
>   configure check ["config file"]
> 
>     Read and parse given config file, but do not use it. useful for
>     checking syntactic and some semantic validity of an config file.
> 
> So you'd want something like "configure check ${CONF_FILE}" I guess.
> That way if you ever have two instances of bird running, reloading the
> second instance won't check the config file for the first instance.

You’re right, I added it:
	BIRD_CHECK_CONF="birdc ${client_args} configure check \"${CONF_FILE}\""

-- 
Alarig

Re: [gentoo-user] Question about bird init script

[Alarig Le Lay]

Hi!

On mar.  5 mars 10:01:08 2019, Michael Orlitzky wrote:
> Once you've used it for a while and are confident that everything works,
> please file a bug to include the improved script in the tree!

All my bird instances are running with my script and I didn’t run in any
issue so far, so I opened a bug to propose it:
https://bugs.gentoo.org/682628

-- 
Alarig