ping_ plugin still use ping6

[solsticedhiver]

ping6 is not available on archlinux (has it ever been ?)

so the /usr/lib/munin/plugins/ping_ plugin of munin-node fails when using ipv6 like in /etc/munin/plugins/ping6_2606:4700:4700::1111

# munin-run "ping6_2606:4700:4700::1111"
/etc/munin/plugins/ping6_2606:4700:4700::1111: line 94: ping6: command not found

the fix is simple and it needs a one character patch:

--- /usr/lib/munin/plugins/ping_ 2019-12-07 20:04:17.263575467 +0100
+++ /usr/lib/munin/plugins/ping_ 2019-12-07 19:56:01.635908455 +0100
@@ -59,7 +59,7 @@

case $0 in
*ping6_*)
- PING=ping6
+ PING=ping
file_host=${0##*/ping6_}
V=IPv6
;;

[je3kmz]

The commands are different between the BSD system and Linux.

--- ./plugins/node.d/ping_.in.orig      2019-10-17 23:49:12.000000000 +0000
+++ ./plugins/node.d/ping_.in   2019-12-10 02:58:21.897377000 +0000
@@ -57,14 +57,14 @@ Foundation, Inc., 51 Franklin Street, Fi
 # shellcheck disable=SC1090
 . "$MUNIN_LIBDIR/plugins/plugin.sh"

+PING=ping
 case $0 in
     *ping6_*)
-        PING=ping6
+       PING=`( uname -a | grep -q 'BSD' ) && echo ping6`
         file_host=${0##*/ping6_}
         V=IPv6
         ;;
     *ping_*)
-        PING=ping
         file_host=${0##*/ping_}
         V=IPv4
         ;;

[solsticedhiver]

Ok.

But the command is empty on anything else than BSD. May be:

PING="ping`( echo BSD | grep -q 'BSD' ) && echo 6`"

[je3kmz]

You've found it well.

PING="ping`( uname -a | grep -q 'BSD' ) && echo 6`"

[solsticedhiver]

Moreover, if ever you are loosing internet connectivity (or the link to the machine you ping), and that ping never returns nor timeout, munin-node waits indefinitly (or some amount of time) that will block the other script run on the node.

Yesterday, I lost internet, but ping never returned and was waiting, waiting... alll the other munin plugins of the node did not run because of the stalled ping_ plugin. for hours.

so I added -W 60 to the ping option. This should definitvly be a default.

Moreover, I think this is a bug of munin-node. Each plugin should be run inside a command similar to timeout(1), and the plugin should be killed if it times out after some predefined time.

[sumpfralle]

Yesterday, I lost internet, but ping never returned and was waiting, waiting... alll the other munin plugins of the node did not run because of the stalled ping_ plugin. for hours.

This should not happen. Every plugin is executed in (see do_with_timeout in Munin::Common::Timeout). A quick test on my host indeed kills the hanging plugin (I just added a long sleep to it) and returns in time.

Please open a separate issue and describe your specific environment, if you are able to reproduce a hanging plugin.

[solsticedhiver]

It was a guess. Because it happenned at the same time I lost internet connectivity.

I am not sure I can reproduce that. It happenned only once, and it stopped as soon as internet connectivity was there again. I did not had time to investigate

[shtrom]

You could use ping_2606:4700:4700::1111 as the symlink. If using a DNS name, you could add env.ping_args -6 to the config to force IPv6 resolution (the only remaining issue would be the title of the graph, showing IPv4).

[solsticedhiver]

This seems to have been fixed in 2.0.61. However there is a new bug ! #1325

[sumpfralle]

Correct, this specific problem should be fixed now, but the which approach does not seem to be as platform-independent, as I thought.

Let's close this specific issue for now and track the issue with which in #1325.