[kubernetes] add kubelogs option

If your cluster is completely broken you can't gather
logs using kubectl / the API, so add an option to easily
copy logs from the most critical pods.

Signed-off-by: Etienne Champetier <e.champetier@ateme.com>

sos/report/plugins/kubernetes.py

@@ -53,6 +53,12 @@ class Kubernetes(Plugin):
     forbidden_paths = [
         "/etc/kubernetes/pki",
     ]
+    kube_system_logs = [
+        "/var/log/pods/kube-system_etcd-*",
+        "/var/log/pods/kube-system_kube-apiserver-*",
+        "/var/log/pods/kube-system_kube-controller-manager-*",
+        "/var/log/pods/kube-system_kube-scheduler-*",
+    ]
     resources = [
         'events',
         'deployments',
@@ -88,8 +94,10 @@ class Kubernetes(Plugin):
                   desc='collect all namespace output separately'),
         PluginOpt('describe', default=False,
                   desc='collect describe output of all resources'),
+        PluginOpt('kubelogs', default=False,
+                  desc='copy some kube-system pod logs without using the API'),
         PluginOpt('podlogs', default=False,
-                  desc='capture stdout/stderr logs from pods'),
+                  desc='capture stdout/stderr logs from pods using the API'),
         PluginOpt('podlogs-filter', default='', val_type=str,
                   desc='only collect logs from pods matching this pattern')
     ]
@@ -113,6 +121,9 @@ def setup(self):
 
         self.add_forbidden_path(self.forbidden_paths)
 
+        if self.get_option('kubelogs'):
+            self.add_copy_spec(self.kube_system_logs)
+
         self.add_env_var([
             'KUBECONFIG',
             'KUBERNETES_HTTP_PROXY',