diff --git a/cluster/log-dump.sh b/cluster/log-dump.sh
new file mode 100755
index 00000000000..630bd3f4895
--- /dev/null
+++ b/cluster/log-dump.sh
@@ -0,0 +1,95 @@
+#!/bin/bash
+
+# Copyright 2016 The Kubernetes Authors All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Call this to dump all master and node logs into the folder specified in $1
+# (defaults to _artifacts). Only works if the provider supports SSH.
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+KUBE_ROOT=$(dirname "${BASH_SOURCE}")/..
+: ${KUBE_CONFIG_FILE:="config-test.sh"}
+
+source "${KUBE_ROOT}/cluster/kube-env.sh"
+source "${KUBE_ROOT}/cluster/kube-util.sh"
+
+readonly report_dir="${1:-_artifacts}"
+echo "Dumping master and node logs to ${report_dir}"
+
+# Saves a single output of running a given command ($2) on a given node ($1)
+# into a given local file ($3). Does not fail if the ssh command fails for any
+# reason, just prints an error to stderr.
+function save-log() {
+    local -r node_name="${1}"
+    local -r cmd="${2}"
+    local -r output_file="${3}"
+    if ! ssh-to-node "${node_name}" "${cmd}" > "${output_file}"; then
+        echo "${cmd} failed for ${node_name}" >&2
+    fi
+}
+
+# Saves logs common to master and nodes. The node name is in $1 and the
+# directory/name prefix is in $2. Assumes KUBERNETES_PROVIDER is set.
+function save-common-logs() {
+    local -r node_name="${1}"
+    local -r prefix="${2}"
+    echo "Dumping logs for ${node_name}"
+    save-log "${node_name}" "cat /var/log/kern.log" "${prefix}-kern.log"
+    save-log "${node_name}" "cat /var/log/docker.log" "${prefix}-docker.log"
+    if [[ "${KUBERNETES_PROVIDER}" == "gce" ]]; then
+        save-log "${node_name}" "cat /var/log/startupscript.log" "${prefix}-startupscript.log"
+    fi
+    if ssh-to-node "${node_name}" "sudo systemctl status kubelet.service" &> /dev/null; then
+        save-log "${node_name}" "sudo journalctl --output=cat -u kubelet.service" "${prefix}-kubelet.log"
+    else
+        save-log "${node_name}" "cat /var/log/kubelet.log" "${prefix}-kubelet.log"
+        save-log "${node_name}" "cat /var/log/supervisor/supervisord.log" "${prefix}-supervisord.log"
+        save-log "${node_name}" "cat /var/log/supervisor/kubelet-stdout.log" "${prefix}-supervisord-kubelet-stdout.log"
+        save-log "${node_name}" "cat /var/log/supervisor/kubelet-stderr.log" "${prefix}-supervisord-kubelet-stderr.log"
+    fi
+}
+
+readonly master_ssh_supported_providers="gce aws kubemark"
+readonly node_ssh_supported_providers="gce gke aws"
+
+if [[ ! "${master_ssh_supported_providers}" =~ "${KUBERNETES_PROVIDER}" ]]; then
+    echo "Master SSH not supported for ${KUBERNETES_PROVIDER}"
+elif ! $(detect-master &> /dev/null); then
+    echo "Master not detected. Is the cluster up?"
+else
+    echo "Master Name: ${MASTER_NAME}"
+    readonly master_prefix="${report_dir}/${MASTER_NAME}"
+    save-log "${MASTER_NAME}" "cat /var/log/kube-apiserver.log" "${master_prefix}-kube-apiserver.log"
+    save-log "${MASTER_NAME}" "cat /var/log/kube-scheduler.log" "${master_prefix}-kube-scheduler.log"
+    save-log "${MASTER_NAME}" "cat /var/log/kube-controller-manager.log" "${master_prefix}-kube-controller-manager.log"
+    save-log "${MASTER_NAME}" "cat /var/log/etcd.log" "${master_prefix}-kube-etcd.log"
+    save-common-logs "${MASTER_NAME}" "${master_prefix}"
+fi
+
+detect-node-names &> /dev/null
+if [[ ! "${node_ssh_supported_providers}"  =~ "${KUBERNETES_PROVIDER}" ]]; then
+    echo "Node SSH not supported for ${KUBERNETES_PROVIDER}"
+elif [[ "${#NODE_NAMES[@]}" -eq 0 ]]; then
+    echo "Nodes not detected. Is the cluster up?"
+else
+    echo "Node Names: ${NODE_NAMES[*]}"
+    for node_name in "${NODE_NAMES[@]}"; do
+        node_prefix="${report_dir}/${node_name}"
+        save-log "${node_name}" "cat /var/log/kube-proxy.log" "${node_prefix}-kube-proxy.log"
+        save-common-logs "${node_name}" "${node_prefix}"
+    done
+fi
diff --git a/hack/jenkins/e2e-runner.sh b/hack/jenkins/e2e-runner.sh
index 5834451df0b..997ea60c0a5 100755
--- a/hack/jenkins/e2e-runner.sh
+++ b/hack/jenkins/e2e-runner.sh
@@ -193,7 +193,15 @@ if [[ "${gcp_list_resources}" == "true" ]]; then
   ${gcp_list_resources_script} > "${gcp_resources_before}"
 fi
 if [[ "${E2E_UP,,}" == "true" ]]; then
-    go run ./hack/e2e.go ${E2E_OPT:-} -v --up
+    # We want to try to gather logs even if kube-up fails, so collect the
+    # result here and fail after dumping logs if it's nonzero.
+    go run ./hack/e2e.go ${E2E_OPT:-} -v --up && up_result="$?" || up_result="$?"
+    if [[ "${up_result}" -ne 0 ]]; then
+        if [[ -x "cluster/log-dump.sh"  ]]; then
+            ./cluster/log-dump.sh "${ARTIFACTS}"
+        fi
+        exit "${up_result}"
+    fi
     go run ./hack/e2e.go -v --ctl="version --match-server-version=false"
     if [[ "${gcp_list_resources}" == "true" ]]; then
       ${gcp_list_resources_script} > "${gcp_resources_cluster_up}"
diff --git a/hack/jenkins/job-configs/global.yaml b/hack/jenkins/job-configs/global.yaml
index 179f50213a0..9fc4f226fb9 100644
--- a/hack/jenkins/job-configs/global.yaml
+++ b/hack/jenkins/job-configs/global.yaml
@@ -11,9 +11,9 @@
 - publisher:
     name: gcs-uploader
     publishers:
-        # Use our script for build artifacts, since it's more flexible.
         - postbuildscript:
             builders:
+                # Use our script for build artifacts, since it's more flexible.
                 - shell: |
                     if [[ -x ./hack/jenkins/upload-to-gcs.sh ]]; then
                       ./hack/jenkins/upload-to-gcs.sh
diff --git a/hack/verify-flags/exceptions.txt b/hack/verify-flags/exceptions.txt
index d7e41963144..421b1e258c6 100644
--- a/hack/verify-flags/exceptions.txt
+++ b/hack/verify-flags/exceptions.txt
@@ -5,7 +5,6 @@ cluster/aws/templates/configure-vm-aws.sh:  # We set the hostname_override to th
 cluster/aws/templates/configure-vm-aws.sh:  api_servers: '${API_SERVERS}'
 cluster/aws/templates/configure-vm-aws.sh:  env-to-grains "hostname_override"
 cluster/aws/templates/configure-vm-aws.sh:  env-to-grains "runtime_config"
-cluster/aws/templates/salt-minion.sh:# We set the hostname_override to the full EC2 private dns name
 cluster/centos/util.sh:  local node_ip=${node#*@}
 cluster/gce/configure-vm.sh:  advertise_address: '${EXTERNAL_IP}'
 cluster/gce/configure-vm.sh:  api_servers: '${KUBERNETES_MASTER_NAME}'
@@ -50,6 +49,10 @@ cluster/juju/charms/trusty/kubernetes/hooks/network-relation-changed:    for k i
 cluster/juju/charms/trusty/kubernetes/hooks/network-relation-changed:    if api_servers:
 cluster/lib/logging.sh:      local source_file=${BASH_SOURCE[$frame_no]}
 cluster/lib/logging.sh:  local source_file=${BASH_SOURCE[$stack_skip]}
+cluster/log-dump.sh:    for node_name in "${NODE_NAMES[@]}"; do
+cluster/log-dump.sh:    local -r node_name="${1}"
+cluster/log-dump.sh:    local -r node_name="${1}"
+cluster/log-dump.sh:readonly report_dir="${1:-_artifacts}"
 cluster/mesos/docker/km/build.sh:  km_path=$(find-binary km darwin/amd64)
 cluster/rackspace/util.sh:    local node_ip=$(nova show --minimal ${NODE_NAMES[$i]} \
 cluster/saltbase/salt/kube-addons/kube-addons.sh:# Create admission_control objects if defined before any other addon services. If the limits
@@ -81,7 +84,6 @@ docs/getting-started-guides/coreos/azure/lib/deployment_logic/kubernetes.js:  re
 docs/getting-started-guides/coreos/azure/lib/deployment_logic/kubernetes.js:  return cloud_config.process_template(input_file, output_file, function(data) {
 docs/getting-started-guides/coreos/azure/lib/deployment_logic/kubernetes.js:  var write_files_extra = cloud_config.write_files_from('addons', '/etc/kubernetes/addons');
 docs/getting-started-guides/coreos/azure/lib/deployment_logic/kubernetes.js:var cloud_config = require('../cloud_config.js');
-docs/getting-started-guides/docker-multinode/skydns-rc.yaml.in:        - -kube_master_url=http://{kube_server_url}:8080
 examples/cluster-dns/images/frontend/client.py:  service_address = socket.gethostbyname(hostname)
 examples/vitess/env.sh:    node_ip=$(get_node_ip)
 hack/jenkins/job-builder-image/Dockerfile:# JJB configuration lives in /etc/jenkins_jobs/jenkins_jobs.ini
@@ -91,11 +93,10 @@ hack/jenkins/update-jobs.sh:  # jenkins_jobs.ini contains administrative credent
 hack/jenkins/update-jobs.sh:  if [[ -e jenkins_jobs.ini ]]; then
 hack/local-up-cluster.sh:      runtime_config="--runtime-config=${RUNTIME_CONFIG}"
 hack/local-up-cluster.sh:    runtime_config=""
+pkg/kubelet/network/hairpin/hairpin.go:	hairpinModeRelativePath = "hairpin_mode"
 pkg/kubelet/qos/memory_policy_test.go:			t.Errorf("oom_score_adj should be between %d and %d, but was %d", test.lowOOMScoreAdj, test.highOOMScoreAdj, oomScoreAdj)
 pkg/kubelet/qos/memory_policy_test.go:	highOOMScoreAdj int // The min oom_score_adj score the container should be assigned.
 pkg/kubelet/qos/memory_policy_test.go:	lowOOMScoreAdj  int // The max oom_score_adj score the container should be assigned.
-pkg/util/oom/oom_linux.go:			err = fmt.Errorf("failed to read oom_score_adj: %v", readErr)
-pkg/util/oom/oom_linux.go:			err = fmt.Errorf("failed to set oom_score_adj to %d: %v", oomScoreAdj, writeErr)
 pkg/util/oom/oom_linux.go:		return fmt.Errorf("invalid PID %d specified for oom_score_adj", pid)
 pkg/util/oom/oom_linux.go:	oomScoreAdjPath := path.Join("/proc", pidStr, "oom_score_adj")
 pkg/util/oom/oom_linux.go:// Writes 'value' to /proc/<pid>/oom_score_adj for all processes in cgroup cgroupName.
@@ -110,4 +111,3 @@ test/e2e/host_path.go:			fmt.Sprintf("--retry_time=%d", retryDuration),
 test/images/mount-tester/mt.go:	flag.BoolVar(&breakOnExpectedContent, "break_on_expected_content", true, "Break out of loop on expected content, (use with --file_content_in_loop flag only)")
 test/images/mount-tester/mt.go:	flag.IntVar(&retryDuration, "retry_time", 180, "Retry time during the loop")
 test/images/mount-tester/mt.go:	flag.StringVar(&readFileContentInLoopPath, "file_content_in_loop", "", "Path to read the file content in loop from")
-pkg/kubelet/network/hairpin/hairpin.go:	hairpinModeRelativePath = "hairpin_mode"