Seed an empty /nsm/kernelrepo so the manager repo is always valid

so-repo-sync only populates /nsm/kernelrepo after the highstate, so on a
manager the file:///nsm/kernelrepo repo could be assigned before any
repodata exists, failing every dnf op. Run createrepo on the dir when
repodata/repomd.xml is missing, leaving a synced repo untouched.

salt/elasticfleet/config.sls

@@ -173,7 +173,7 @@ eaoptionalintegrationsdir:
 
 {% for minion in node_data %}
 {% set role = node_data[minion]["role"] %}
-{% if role in [ "eval","fleet","import","manager", "managerhype", "managersearch","standalone" ] %}
+{% if role in [ "eval","fleet","heavynode","import","manager", "managerhype", "managersearch","standalone" ] %}
 {% set optional_integrations = ELASTICFLEETMERGED.optional_integrations %}
 {% set integration_keys = optional_integrations.keys() %}
 fleet_server_integrations_{{ minion }}:

salt/elasticfleet/manager.sls

@@ -67,6 +67,8 @@ so-elastic-fleet-package-upgrade:
         interval: 30
     - require:
       - http: wait_for_so-kibana
+    - onchanges:
+      - file: /opt/so/state/elastic_fleet_packages.txt
 
 so-elastic-fleet-integrations:
   cmd.run:

salt/elasticfleet/tools/sbin/so-elastic-fleet-integration-policy-load

@@ -9,11 +9,13 @@
 RETURN_CODE=0
 
 if [ ! -f /opt/so/state/eaintegrations.txt ]; then
+  # First, check for any package upgrades
+  /usr/sbin/so-elastic-fleet-package-upgrade
 
-  # update Fleet Server policies
+  # Second, update Fleet Server policies
   /usr/sbin/so-elastic-fleet-integration-policy-elastic-fleet-server
 
-  # configure Elastic Defend Integration separately
+  # Third, configure Elastic Defend Integration seperately
   /usr/sbin/so-elastic-fleet-integration-policy-elastic-defend
 
   # Each group fetches its agent policy once and dispatches create/update writes concurrently.
@@ -30,12 +32,9 @@ if [ ! -f /opt/so/state/eaintegrations.txt ]; then
   elastic_fleet_load_integrations_dir "so-grid-nodes_heavy" \
     /opt/so/conf/elastic-fleet/integrations/grid-nodes_heavy "Grid Nodes Policy_Heavy" || RETURN_CODE=1
 
-  # Fleet Server - Optional integrations (adds integration configuration to a given FleetServer_ policy)
+  # Fleet Server - Optional integrations (one agent policy per FleetServer_* directory)
   for FLEET_DIR in /opt/so/conf/elastic-fleet/integrations-optional/FleetServer*/; do
     [ -d "$FLEET_DIR" ] || continue
-    INTEGRATIONS=("${FLEET_DIR%/}"/*.json)
-    [ -e "${INTEGRATIONS[0]}" ] || continue
-
     FLEET_POLICY=$(basename "$FLEET_DIR")
     elastic_fleet_load_integrations_dir "$FLEET_POLICY" \
       "${FLEET_DIR%/}" "Fleet Server Policy" "elasticsearch-logs" || RETURN_CODE=1

salt/elasticfleet/tools/sbin_jinja/so-elastic-fleet-package-upgrade

@@ -12,22 +12,17 @@ PKG_LOAD_FAILURES=0
 PKG_LOAD_FAILURES_NAMES=()
 
 {%- for PACKAGE in SUPPORTED_PACKAGES %}
-if INSTALLED_VERSION=$(elastic_fleet_package_version_check "{{ PACKAGE }}") && LATEST_VERSION=$(elastic_fleet_package_latest_version_check "{{ PACKAGE }}"); then
+echo "Upgrading {{ PACKAGE }} package..."
-
+if VERSION=$(elastic_fleet_package_latest_version_check "{{ PACKAGE }}"); then
-    if [ "$INSTALLED_VERSION" == "$LATEST_VERSION" ]; then
+    if ! elastic_fleet_package_install "{{ PACKAGE }}" "$VERSION"; then
-        echo "{{ PACKAGE }} integration version $INSTALLED_VERSION is already at the reported latest version $LATEST_VERSION, skipping upgrade."
+        PKG_LOAD_FAILURES=$((PKG_LOAD_FAILURES + 1))
-    else
+        PKG_LOAD_FAILURES_NAMES+=("{{ PACKAGE }}")
-        echo "Upgrading {{ PACKAGE }} package to version $LATEST_VERSION..."
-        if ! elastic_fleet_package_install "{{ PACKAGE }}" "$LATEST_VERSION"; then
-            PKG_LOAD_FAILURES=$((PKG_LOAD_FAILURES + 1))
-            PKG_LOAD_FAILURES_NAMES+=("{{ PACKAGE }}")
-        fi
     fi
 else
-    echo "ERROR: Failed to get version information for integration {{ PACKAGE }}"
     PKG_LOAD_FAILURES=$((PKG_LOAD_FAILURES + 1))
     PKG_LOAD_FAILURES_NAMES+=("{{ PACKAGE }}")
 fi
+echo
 {%- endfor %}
 
 if [ $PKG_LOAD_FAILURES -gt 0 ]; then
@@ -40,3 +35,6 @@ if [ $PKG_LOAD_FAILURES -gt 0 ]; then
 else
     echo "Successfully upgraded all packages."
 fi
+
+echo
+/usr/sbin/so-elasticsearch-templates-load

salt/elasticfleet/tools/sbin_jinja/so-elastic-fleet-setup

@@ -181,9 +181,6 @@ if ! elastic_fleet_policy_create "so-grid-nodes_heavy" "SO Grid Nodes - Heavy No
     exit 1
 fi
 
-# Check for package upgrades
-so-elastic-fleet-package-upgrade
-
 # Load Integrations for default policies
 so-elastic-fleet-integration-policy-load
 

salt/manager/files/mirror-kernel.txt

@@ -0,0 +1,2 @@
+https://repo.securityonion.net/file/so-repo/prod/3/oracle/9-uek8
+https://repo-alt.securityonion.net/prod/3/oracle/9-uek8

salt/manager/files/repodownload.conf

@@ -10,4 +10,9 @@ keepcache=0
 name=Security Onion Repo repo
 mirrorlist=file:///opt/so/conf/reposync/mirror.txt
 enabled=1
-gpgcheck=1
+gpgcheck=1
+[securityonionkernel]
+name=Security Onion Repo repo
+mirrorlist=file:///opt/so/conf/reposync/mirror-kernel.txt
+enabled=1
+gpgcheck=1

salt/manager/init.sls

@@ -86,6 +86,28 @@ repo_dir:
       - group
     - show_changes: False
 
+kernelrepo_dir:
+  file.directory:
+    - name: /nsm/kernelrepo
+    - user: socore
+    - group: socore
+    - recurse:
+      - user
+      - group
+    - show_changes: False
+
+# Ensure /nsm/kernelrepo is always a valid (if empty) repo before it is ever assigned to
+# a client. Without repodata/repomd.xml an enabled file:///nsm/kernelrepo repo makes every
+# dnf operation fail; so-repo-sync only populates it after the highstate, so seed an empty
+# repo here. Only runs when repodata is missing, so it won't clobber a synced repo.
+kernelrepo_init_empty:
+  cmd.run:
+    - name: createrepo /nsm/kernelrepo
+    - unless: 'test -e /nsm/kernelrepo/repodata/repomd.xml'
+    - require:
+      - file: kernelrepo_dir
+      - pkg: install_createrepo
+
 manager_sbin:
   file.recurse:
     - name: /usr/sbin
@@ -122,6 +144,13 @@ so-repo-mirrorlist:
     - user: socore
     - group: socore
 
+so-repo-kernel-mirrorlist:
+  file.managed:
+    - name: /opt/so/conf/reposync/mirror-kernel.txt
+    - source: salt://manager/files/mirror-kernel.txt
+    - user: socore
+    - group: socore
+
 so-repo-sync:
   {%     if MANAGERMERGED.reposync.enabled %}
   cron.present:

salt/manager/tools/sbin/so-repo-sync

@@ -10,5 +10,16 @@ NOROOT=1
 set -e
 
 curl --retry 5 --retry-delay 60 -A "reposync/$(sync_options)" https://sigs.securityonion.net/checkup --output /tmp/checkup
+
 dnf reposync --norepopath -g --delete -m -c /opt/so/conf/reposync/repodownload.conf --repoid=securityonionsync --download-metadata -p /nsm/repo/
 createrepo /nsm/repo
+
+# The kernel repo section is deployed to repodownload.conf by the manager highstate, which
+# runs AFTER this script during soup. On the first upgrade to a kernel-aware version the
+# on-disk config still predates the section, so guard on its presence to avoid dnf's
+# "Unknown repo: 'securityonionkernel'" aborting the sync (set -e). The next sync after the
+# highstate deploys the section will pick it up.
+if grep -q '^\[securityonionkernel\]' /opt/so/conf/reposync/repodownload.conf; then
+  dnf reposync --norepopath -g --delete -m -c /opt/so/conf/reposync/repodownload.conf --repoid=securityonionkernel --download-metadata -p /nsm/kernelrepo/
+  createrepo /nsm/kernelrepo
+fi

salt/nginx/enabled.sls

@@ -59,6 +59,7 @@ so-nginx:
       - /opt/so/conf/navigator/layers/:/opt/socore/html/navigator/assets/so:ro
       - /opt/so/conf/navigator/config.json:/opt/socore/html/navigator/assets/config.json:ro
       - /nsm/repo:/opt/socore/html/repo:ro
+      - /nsm/kernelrepo:/opt/socore/html/kernelrepo:ro
       - /nsm/rules:/nsm/rules:ro
       {%   if NGINXMERGED.external_suricata %}
       - /opt/so/rules/nids/suri:/surirules:ro

salt/repo/client/oracle.sls

@@ -57,6 +57,26 @@ so_repo:
     - enabled: 1
     - gpgcheck: 1
 
+so_kernel_repo:
+  pkgrepo.managed:
+    - name: securityonionkernel
+    - humanname: Security Onion Kernel Repo
+  {% if GLOBALS.is_manager %}
+    - baseurl: file:///nsm/kernelrepo/
+  {% else %}
+    - baseurl: https://{{ GLOBALS.repo_host }}/kernelrepo
+  {% endif %}
+    - enabled: 1
+    - gpgcheck: 1
+    # Supplementary kernel repo: tolerate it being empty/unreachable (e.g. before the
+    # manager has populated /nsm/kernelrepo) so a missing repomd.xml can't make every
+    # dnf/pkg operation on the grid fail.
+    - skip_if_unavailable: 1
+    # Only assign the kernel repo once physical NIC names are pinned by MAC, so the
+    # UEK8 kernel update can't renumber interfaces SO binds by name (see pin_nic_names
+    # in salt/common/init.sls, which drops this marker via /usr/sbin/so-nic-pin).
+    - onlyif: 'test -e /opt/so/state/nic_names_pinned'
+
 {% endif %}
   
 # TODO: Add a pillar entry for custom repos

setup/so-functions

@@ -886,6 +886,7 @@ create_repo() {
 	title "Create the repo directory"
 	logCmd "dnf -y install yum-utils createrepo_c"
 	logCmd "createrepo /nsm/repo"
+	logCmd "createrepo /nsm/kernelrepo"
 }
 
 
@@ -1812,6 +1813,16 @@ securityonion_repo() {
 			echo "mirrorlist=file:///etc/yum/mirror.txt" >> /etc/yum.repos.d/securityonion.repo
 			echo "enabled=1" >> /etc/yum.repos.d/securityonion.repo
 			echo "gpgcheck=1" >> /etc/yum.repos.d/securityonion.repo
+			echo "https://repo.securityonion.net/file/so-repo/prod/3/oracle/9-uek8" > /etc/yum/mirror-kernel.txt
+			echo "https://so-repo-east.s3.us-east-005.backblazeb2.com/prod/3/oracle/9-uek8" >> /etc/yum/mirror-kernel.txt
+			echo "[securityonionkernel]" > /etc/yum.repos.d/securityonionkernel.repo
+			echo "name=Security Onion Kernel Repo repo" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "mirrorlist=file:///etc/yum/mirror-kernel.txt" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "enabled=1" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "gpgcheck=1" >> /etc/yum.repos.d/securityonionkernel.repo
+			# Supplementary kernel repo: tolerate it being empty/unreachable so a missing
+			# repomd.xml can't make every dnf operation fail before the repo is populated.
+			echo "skip_if_unavailable=1" >> /etc/yum.repos.d/securityonionkernel.repo
 			logCmd "dnf repolist"
 		else
 			echo "[securityonion]" > /etc/yum.repos.d/securityonion.repo
@@ -1820,6 +1831,13 @@ securityonion_repo() {
 			echo "enabled=1" >> /etc/yum.repos.d/securityonion.repo
 			echo "gpgcheck=1" >> /etc/yum.repos.d/securityonion.repo
 			echo "sslverify=0" >> /etc/yum.repos.d/securityonion.repo
+			echo "[securityonionkernel]" > /etc/yum.repos.d/securityonionkernel.repo
+			echo "name=Security Onion Kernel Repo" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "baseurl=https://$MSRV/kernelrepo" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "enabled=1" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "gpgcheck=1" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "sslverify=0" >> /etc/yum.repos.d/securityonionkernel.repo
+			echo "skip_if_unavailable=1" >> /etc/yum.repos.d/securityonionkernel.repo
 			logCmd "dnf repolist"
 		fi
 	elif [[ ! $waitforstate ]]; then
@@ -1829,12 +1847,25 @@ securityonion_repo() {
 		echo "enabled=1" >> /etc/yum.repos.d/securityonion.repo
 		echo "gpgcheck=1" >> /etc/yum.repos.d/securityonion.repo
 		echo "sslverify=0" >> /etc/yum.repos.d/securityonion.repo
+		echo "[securityonionkernel]" > /etc/yum.repos.d/securityonionkernel.repo
+		echo "name=Security Onion Kernel Repo" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "baseurl=https://$MSRV/kernelrepo" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "enabled=1" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "gpgcheck=1" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "sslverify=0" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "skip_if_unavailable=1" >> /etc/yum.repos.d/securityonionkernel.repo
 	elif [[ $waitforstate ]]; then
 		echo "[securityonion]" > /etc/yum.repos.d/securityonion.repo
 		echo "name=Security Onion Repo" >> /etc/yum.repos.d/securityonion.repo
 		echo "baseurl=file:///nsm/repo/" >> /etc/yum.repos.d/securityonion.repo
 		echo "enabled=1" >> /etc/yum.repos.d/securityonion.repo
 		echo "gpgcheck=1" >> /etc/yum.repos.d/securityonion.repo
+		echo "[securityonionkernel]" > /etc/yum.repos.d/securityonionkernel.repo
+		echo "name=Security Onion Kernel Repo" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "baseurl=file:///nsm/kernelrepo/" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "enabled=1" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "gpgcheck=1" >> /etc/yum.repos.d/securityonionkernel.repo
+		echo "skip_if_unavailable=1" >> /etc/yum.repos.d/securityonionkernel.repo
 	fi
 	logCmd "dnf repolist all"
 	if [[ $waitforstate ]]; then
@@ -1850,9 +1881,12 @@ repo_sync_local() {
 	# Sync the repo from the SO repo locally.
 	info "Adding Repo Download Configuration"
 	mkdir -p /nsm/repo
+	mkdir -p /nsm/kernelrepo
 	mkdir -p /opt/so/conf/reposync/cache
 	echo "https://repo.securityonion.net/file/so-repo/prod/3/oracle/9" > /opt/so/conf/reposync/mirror.txt
 	echo "https://repo-alt.securityonion.net/prod/3/oracle/9" >> /opt/so/conf/reposync/mirror.txt
+	echo "https://repo.securityonion.net/file/so-repo/prod/3/oracle/9-uek8" > /opt/so/conf/reposync/mirror-kernel.txt
+	echo "https://repo-alt.securityonion.net/prod/3/oracle/9-uek8" >> /opt/so/conf/reposync/mirror-kernel.txt
 	echo "[main]" > /opt/so/conf/reposync/repodownload.conf
 	echo "gpgcheck=1" >> /opt/so/conf/reposync/repodownload.conf
 	echo "installonly_limit=3" >> /opt/so/conf/reposync/repodownload.conf
@@ -1866,12 +1900,18 @@ repo_sync_local() {
 	echo "mirrorlist=file:///opt/so/conf/reposync/mirror.txt" >> /opt/so/conf/reposync/repodownload.conf
 	echo "enabled=1" >> /opt/so/conf/reposync/repodownload.conf
 	echo "gpgcheck=1" >> /opt/so/conf/reposync/repodownload.conf
+	echo "[securityonionkernel]" >> /opt/so/conf/reposync/repodownload.conf
+	echo "name=Security Onion Kernel Repo repo" >> /opt/so/conf/reposync/repodownload.conf
+	echo "mirrorlist=file:///opt/so/conf/reposync/mirror-kernel.txt" >> /opt/so/conf/reposync/repodownload.conf
+	echo "enabled=1" >> /opt/so/conf/reposync/repodownload.conf
+	echo "gpgcheck=1" >> /opt/so/conf/reposync/repodownload.conf
 
 	logCmd "dnf repolist"
 
 	if [[ ! $is_airgap ]]; then
 		curl --retry 5 --retry-delay 60 -A "netinstall/$SOVERSION/$OS/$(uname -r)/1" https://sigs.securityonion.net/checkup --output /tmp/install
 		retry 5 60 "dnf reposync --norepopath -g --delete -m -c /opt/so/conf/reposync/repodownload.conf --repoid=securityonionsync --download-metadata -p /nsm/repo/" >> "$setup_log" 2>&1 || fail_setup
+		retry 5 60 "dnf reposync --norepopath -g --delete -m -c /opt/so/conf/reposync/repodownload.conf --repoid=securityonionkernel --download-metadata -p /nsm/kernelrepo/" >> "$setup_log" 2>&1 || fail_setup
 		# After the download is complete run createrepo
 		create_repo
 	fi
@@ -2228,6 +2268,13 @@ update_sudoers_for_testing() {
 }
 
 update_packages() {
+	# Pin physical NIC names by MAC BEFORE pulling packages, so the UEK8 kernel that
+	# the update below installs can't renumber the interfaces SO binds by name. Doing
+	# it here (instead of waiting for the common highstate) also drops the
+	# /opt/so/state/nic_names_pinned marker that gates the kernel repo, so the kernel
+	# repo is assigned on the very first highstate and the kernel isn't downgraded and
+	# then re-upgraded. Run-once: so-nic-pin no-ops if the marker already exists.
+	logCmd "bash ../salt/common/tools/sbin/so-nic-pin"
 	logCmd "dnf repolist"
 	logCmd "dnf -y update --allowerasing --exclude=salt*,docker*,containerd*"
 	RMREPOFILES=("oracle-linux-ol9.repo" "uek-ol9.repo" "virt-ol9.repo")