Cacti not polling every 30 minutes

[jmaris]

Hello,

my Cacti set-up is configured with Spine to poll every 10 seconds. It is currently polling roundabout 6600 DS in approx. 4200 RRD files (quad core CPU).
Via some performance tools like iostat and sar I noticed that the disk I/O is 100% busy, but nominally polling is still OK. (I'm considering the boost plugin to reduce the I/O activity).

However, every 30 minutes sharp, synchronised exactly at the hour and 30 minutes after each hour, Cacti fails to poll during 3 consecutive epochs, so 30 seconds of sampling time lost. I suspect that something is supressing cacti polling.

I cannot find any cron job that runs at these moments in time, nor any process or anything in the log files that could indicate the cause of this (packet?) loss. CPU increases only slightly.

So, I'm wondering whether PHP, MySQL, Apache or Cacti itself are doing something special, or whether I need to look into the OS (Linux Debian 7).

I have also taken a tcpdump around the time and it seems that packets are dropped by the kernel. Outside this 30min instance, tcpdump does not drop any packets.
SSH connectivity remains, so the networking stack is not affected.

Any help or suggestions on how to troubleshoot this further are more than welcome.
Thanks, Jan.

[ylonull]

Hi,

I have a very similar issue (packet loss every 30 minutes), I just wanted to know if you ever made any progress ?