PNArchives/Memoh
1
0
Fork 0
mirror of https://github.com/memohai/Memoh.git synced 2026-04-25 07:00:48 +09:00
Code Issues Packages Projects Releases Wiki Activity

fix(containerd): re-apply CNI network after server restart for running MCP tasks

Browse Source 
Server container restart drops cni0 bridge, veth and iptables masquerade
in its network namespace while MCP tasks keep running in containerd.
Reconcile and ensureContainerAndTask now re-run SetupNetwork for already-
running tasks so outbound connectivity is restored.
This commit is contained in:
BBQ
2026-02-16 16:45:07 +08:00
parent b329475359
commit b8a6a85fbb
1 changed files with 22 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/handlers/containerd.go
+22
View File
@@ -301,6 +301,14 @@ func (h *ContainerdHandler) ensureContainerAndTask(ctx context.Context, containe
}
if len(tasks) > 0 {
if tasks[0].Status == tasktypes.Status_RUNNING {
// Task is running but CNI state may be stale (e.g. server container restarted).
// Re-apply network to ensure connectivity.
if task, taskErr := h.service.GetTask(ctx, containerID); taskErr == nil {
if netErr := ctr.SetupNetwork(ctx, task, containerID); netErr != nil {
h.logger.Warn("network re-setup failed for running task",
slog.String("container_id", containerID), slog.Any("error", netErr))
}
}
return nil
}
if err := h.service.DeleteTask(ctx, containerID, &ctr.DeleteTaskOptions{Force: true}); err != nil {
@@ -928,6 +936,20 @@ func (h *ContainerdHandler) ReconcileContainers(ctx context.Context) {
slog.String("bot_id", botID), slog.Any("error", dbErr))
}
}
// Re-apply CNI networking: server container restart drops cni0 bridge,
// veth endpoints and iptables masquerade rules while the MCP task keeps
// running inside containerd.
if task, taskErr := h.service.GetTask(ctx, containerID); taskErr == nil {
if netErr := ctr.SetupNetwork(ctx, task, containerID); netErr != nil {
h.logger.Warn("reconcile: network re-setup failed for running task",
slog.String("bot_id", botID),
slog.String("container_id", containerID),
slog.Any("error", netErr))
}
} else {
h.logger.Warn("reconcile: failed to get task for network re-setup",
slog.String("bot_id", botID), slog.Any("error", taskErr))
}
h.logger.Info("reconcile: container healthy",
slog.String("bot_id", botID), slog.String("container_id", containerID))
continue