If the TFTP request is able to access the TFTP service pod but is unable to find its way back to the node, it may be because the kernel is not tracking established TFTP connections.
The following image is tcpdump
data from within the TFTP pod. It shows what happens when the TFTP request cannot
find a route back to the node that sent the request. The node IP address is 10.32.0.1
, which is the IP address of the
Kubernetes weave
network. It is forwarding the node’s TFTP to this pod. The server IP address is 10.32.0.13
.
10.32.0.1
, but that
does not have any effect.Check if the nf_nat_tftp
kernel module has been loaded. The kernel module is loaded on all ingress points in the
Kubernetes cluster, so there will likely be no missing kernel modules.
(ncn-mw#
) Load nf_nat_tftp
, if it has not been loaded already.
modprobe nf_nat_tftp
(ncn-mw#
) Restart the cray-tftp
service.
kubectl delete pod cray-tftp-885cc65c4-fk8bm
Example output:
pod "cray-tftp-885cc65c4-fk8bm" deleted