[collectd] loss of data on collectd-server

[eric fauser]

hi again

after a run in debugmode, i found out that also the network-code didn't get 
these packets
(but tcpdump showed them)

>grep apaXoses1 /opt/collectd.log
===
network.c:483:network_receive(): host = apaXoses1, type = cpu, inst = 0, 
value = 1167909667:361893:51:239458:187620033:1497641
network.c:483:network_receive(): host = apaXoses1, type = cpu, inst = 1, 
value = 1167909667:323835:34:247262:187653225:1494781
network.c:483:network_receive(): host = apaXoses1, type = cpu, inst = 0, 
value = 1167909697:361893:51:239458:187622971:1497703
network.c:483:network_receive(): host = apaXoses1, type = cpu, inst = 1, 
value = 1167909697:323835:34:247264:187656207:1494797

so it might be a kernel/buffer problem -> a check with "netstat -su" showed 
in increasing
number of packet receive errors.

>netstat -su
===
Udp:
    153012586 packets received
    533952 packets to unknown port received.
    14478080 packet receive errors
    78870193 packets sent

now i increased the recv-packetbuffer to 8MB and restarted the 
collect-server

>sysctl -w net.core.rmem_default=$((8*1024*1024))
    net.core.rmem_default = 8388608
>sysctl -w net.core.rmem_max=$((8*1024*1024))
    net.core.rmem_max = 8388608

... and the problem has gone ;)

br
eric