[collectd] ping plugin is not works in some cases (Linux)

[Matthias Runge]

On 18/10/2019 08:57, Sergey wrote:

> 
> collectd[1301]: read-function of plugin `ping' failed. Will suspend it for 20.000 seconds.
> collectd[1301]: ping plugin: ping_host_add (localhost) failed: Operation not permitted
> collectd[1301]: ping plugin: No host could be added to ping object. Giving up.
> collectd[1301]: ping plugin: The ping thread had a problem. Restarting it.
> 
> I forgot another difference between working and nonworking systems:
> systemd used as init on nonworking systems. And it's the problem like
> SELinux. It is even mentioned in collectd.service:
> 
> # A few plugins won't work without some privileges, which you'll have to
> # specify using the CapabilityBoundingSet directive below.
> #
> # Here's a (incomplete) list of the plugins known capability requirements:
> #   ceph            CAP_DAC_OVERRIDE
> #   dns             CAP_NET_RAW
> #   exec            CAP_SETUID CAP_SETGID
> #   intel_rdt       CAP_SYS_RAWIO
> #   intel_pmu       CAP_SYS_ADMIN
> #   iptables        CAP_NET_ADMIN
> #   ping            CAP_NET_RAW
> #   processes       CAP_NET_ADMIN  (CollectDelayAccounting only)
> #   smart           CAP_SYS_RAWIO
> #   turbostat       CAP_SYS_RAWIO
> #
> # Example, if you use the iptables plugin alongside the dns or ping plugin:
> #CapabilityBoundingSet=CAP_NET_RAW CAP_NET_ADMIN
> 

Hi,

yes, that sounds plausible. I usually do on my systems:

setsebool -P collectd_tcp_network_connect=on

Another thing which may have hurt you is, that there are two
localhost entries in /etc/hosts, one resolving to 127.0.0.1, one
to the IPv6 address. At least that is something, killing many
ruby applications.

Matthias