ICMP Packet Loss Wrong?

[mooks]

Hi

I've discovered recently a Problem and I'm not shure if Zabbix is the issue or something else.

One of our uplinks has some stability problems, random packet loss.
Therefore, I've reduced the check intervall of Template Module ICMP Ping down to 2s.

Zabbix is shows now very often a packet loss of about 16 - 33,3% on all uplinks of the same ISP.
The uplink checked from an other ISP hardly shows packet loss.

The strange thing is, if i test the uplink over ssh from my zabbix server with ping every 1 sec i have exact 0 packet loss O.o

The queues of Zabbix are all 0 and StartPingers=20

[dimir]

First, I'd not recommend using update interval of 2 seconds, you might flood the target host. Zabbix uses fping utility for icmp checks, you might want to try running it manually from the host where server/proxy is.
Another idea for debugging, you can increase log level of icmp pingers temporarily and see the output of fping:

zabbix_server -Rlog_level_increase="icmp pinger"
tail -F /var/log/zabbix/zabbix_server.log
zabbix_server -Rlog_level_decrease="icmp pinger"

[mooks]

Thank you dimir for your input

Usually I also wouldn't use an interval of 2sec, it's only temporarly.

I've checked the log, nothing special.
However I've noticed that we have some latency spikes to this ISP. It seems like Zabbix is using a much shorter timeout interval, that's why I was not able to reproduce the packet loss using ping

[dimir]

If you're talking about fping -i option then Zabbix is using as little value as possible (0, 1 or 10, depending on fping version). As for -p you can specify it as item parameter. Also I think you could play with number of packets to send when testing.