Surfaced during the PR #852 review (alongside the #823 fix).
The container-log (/var/log/pods) NVMe-relocation block was added to packer/base/install/start_k3s_server.sh / start_k3s_agent.sh — but those files are not in the runtime path (nothing references them; base.pkr.hcl only runs install_k3s.sh). The scripts actually uploaded and run on nodes are src/main/resources/com/rustyrazorblade/easydblab/services/start-k3s-{server,agent}.sh (via K3sService/K3sAgentService), and grep confirms 0 occurrences of /var/log/pods in either live script.
So the pod-log relocation to NVMe is silently not applied on real clusters — pod logs still land on the root volume. (#852 fixed the same dead-file problem for the /mnt/db1 mount drop-in; this log-relocation block is the remaining divergence.)
Fix: port the /var/log/pods relocation into the live start-k3s-*.sh resource scripts, and consider deleting packer/base/install/start_k3s_*.sh to stop this divergence recurring.
Surfaced during the PR #852 review (alongside the #823 fix).
The container-log (
/var/log/pods) NVMe-relocation block was added topacker/base/install/start_k3s_server.sh/start_k3s_agent.sh— but those files are not in the runtime path (nothing references them;base.pkr.hclonly runsinstall_k3s.sh). The scripts actually uploaded and run on nodes aresrc/main/resources/com/rustyrazorblade/easydblab/services/start-k3s-{server,agent}.sh(viaK3sService/K3sAgentService), andgrepconfirms 0 occurrences of/var/log/podsin either live script.So the pod-log relocation to NVMe is silently not applied on real clusters — pod logs still land on the root volume. (#852 fixed the same dead-file problem for the
/mnt/db1mount drop-in; this log-relocation block is the remaining divergence.)Fix: port the
/var/log/podsrelocation into the livestart-k3s-*.shresource scripts, and consider deletingpacker/base/install/start_k3s_*.shto stop this divergence recurring.