I've been troubleshooting issues with high latency on my GKE pods, and a traceroute
yields results like the following when poking the database server for that cluster (which is a GCE VPS, but not part of the cluster):
$ traceroute -n -q5 -w30 10.164.0.2
traceroute to 10.164.0.2 (10.164.0.2), 30 hops max, 60 byte packets
1 10.44.49.1 0.050 ms 0.011 ms 0.007 ms 0.008 ms 0.006 ms
2 * 10.164.0.2 1.700 ms * 1.780 ms *
However, when I traceroute
from the node itself (using toolbox
, since the node is running COS), I get this (which makes sense):
$ traceroute -q5 -w30 10.164.0.2
traceroute to 10.164.0.2 (10.164.0.2), 64 hops max
1 10.164.0.2 1.301ms 0.085ms 0.095ms 0.076ms 0.118ms
What could be/is causing the delays?
Note: I've observed this behaviour on multiple pods spread across different nodes, running both an ancient version of Debian and an up-to-date Alpine.