Merge "Update SEL Events codes to avoid collisions"
This commit is contained in:
Zuul
@@ -11,17 +11,17 @@ log_info() { echo "$(date '+%F %H:%M:%S') INFO: $*"; }
|
||||
log_warn() { echo "$(date '+%F %H:%M:%S') WARN: $*"; }
|
||||
|
||||
declare -Ar CODES=(
|
||||
[bootstrap.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd1"
|
||||
[bootstrap.err]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd2"
|
||||
[config.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd3"
|
||||
[config.err]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd4"
|
||||
[setup.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd5"
|
||||
[setup.err]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd6"
|
||||
[tests.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd7"
|
||||
[tests.err]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd8"
|
||||
[backup.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xd9"
|
||||
[backup.err]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xda"
|
||||
[finished.ok]="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe0"
|
||||
[bootstrap.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xF6"
|
||||
[bootstrap.err]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xF7"
|
||||
[config.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xF8"
|
||||
[config.err]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xF9"
|
||||
[setup.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFA"
|
||||
[setup.err]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFB"
|
||||
[tests.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFC"
|
||||
[tests.err]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFD"
|
||||
[backup.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFE"
|
||||
[backup.err]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xFF"
|
||||
[finished.ok]="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE0"
|
||||
)
|
||||
|
||||
stage="${1:-}"; status="${2:-}"
|
||||
|
||||
@@ -17,6 +17,10 @@ SEED_SERVICE="/etc/systemd/system/cloud-init-seed.service"
|
||||
SEED_NETWORK_CFG="network-config"
|
||||
NETWORK_CFG_FILE="/run/.$SEED_NETWORK_CFG"
|
||||
CLOUD_INIT_IF_FILE="/etc/network/interfaces.d/50-cloud-init"
|
||||
readonly EVENT_FACTORY_SETUP_COMPLETE="factory_setup_complete"
|
||||
readonly EVENT_FACTORY_SETUP_FAILED="factory_setup_failed"
|
||||
readonly DATA_FACTORY_SETUP_COMPLETE="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE0 # \"Factory Setup Complete\""
|
||||
readonly DATA_FACTORY_SETUP_FAILED="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE1 # \"Factory Setup Failed\""
|
||||
|
||||
function check_rc_die {
|
||||
local -i rc=${1}
|
||||
@@ -59,6 +63,32 @@ flock -n 200 || {
|
||||
exit 0
|
||||
}
|
||||
|
||||
function send_ipmi_event {
|
||||
local event_type="$1"
|
||||
local event_data
|
||||
case "$event_type" in
|
||||
"$EVENT_FACTORY_SETUP_COMPLETE") event_data="$DATA_FACTORY_SETUP_COMPLETE" ;;
|
||||
"$EVENT_FACTORY_SETUP_FAILED") event_data="$DATA_FACTORY_SETUP_FAILED" ;;
|
||||
*)
|
||||
log_warn "Unknown IPMI event type: $event_type"
|
||||
return 1
|
||||
;;
|
||||
esac
|
||||
|
||||
local temp_file=$(mktemp /tmp/ipmi_event_XXXXXX.txt)
|
||||
echo "$event_data" > "$temp_file"
|
||||
|
||||
if ipmitool sel add "$temp_file" 2>/dev/null; then
|
||||
log_info "IPMI event sent successfully: $event_type"
|
||||
rm -f "$temp_file"
|
||||
return 0
|
||||
else
|
||||
log_warn "Failed to send IPMI event: $event_type"
|
||||
rm -f "$temp_file"
|
||||
return 1
|
||||
fi
|
||||
}
|
||||
|
||||
# If clean is passed as an argument, remove the udev rule and service,
|
||||
# the custom cloud.cfg file, and the script itself.
|
||||
# This is to ensure that the cloud-init-seed service is not triggered
|
||||
@@ -78,8 +108,10 @@ log_info "Starting cloud-init using seed ISO..."
|
||||
# Checks if factory-install has been completed. This is required to be able
|
||||
# to run cloud-init from a seed ISO.
|
||||
if [[ ! -f "$FACTORY_INSTALL_COMPLETE_FILE" ]]; then
|
||||
log_fatal "Cloud-init from factory-install has not been completed yet. Exiting."
|
||||
send_ipmi_event "$EVENT_FACTORY_SETUP_FAILED"
|
||||
log_fatal "/var/lib/factory-install/stage/complete does not exist. Ensure factory-install was successful."
|
||||
fi
|
||||
send_ipmi_event "$EVENT_FACTORY_SETUP_COMPLETE"
|
||||
|
||||
# Finds the first device found with the label CIDATA or cidata.
|
||||
# If the device is not found, exit the script.
|
||||
|
||||
@@ -72,36 +72,36 @@ readonly EVENT_PLATFORM_CLOUDINIT_UPDATE_COMPLETE="platform_cloudinit_update_com
|
||||
readonly EVENT_PLATFORM_CLOUDINIT_UPDATE_FAILED="platform_cloudinit_update_failed"
|
||||
|
||||
# IPMI payloads (data bytes)
|
||||
readonly DATA_FACTORY_SETUP_COMPLETE="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe0 # \"Factory Setup Complete\""
|
||||
readonly DATA_FACTORY_SETUP_FAILED="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe1 # \"Factory Setup Failed\""
|
||||
readonly DATA_FACTORY_SETUP_COMPLETE="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE0 # \"Factory Setup Complete\""
|
||||
readonly DATA_FACTORY_SETUP_FAILED="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE1 # \"Factory Setup Failed\""
|
||||
|
||||
readonly DATA_APISERVER_CERT_OK="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe2 # \"ApiServer Cert Valid\""
|
||||
readonly DATA_LEAF_CERTS_RENEW_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe3 # \"K8S Leaf Certs Renew Failed\""
|
||||
readonly DATA_RENEW_FAIL_PODS="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe8 # \"Pods Cert Renew Failed\""
|
||||
readonly DATA_CERTMANAGER_CERTS_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xef # \"Cert-manager Secrets Renew Failed\""
|
||||
readonly DATA_RENEW_FAIL_KUBECTL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xf0 # \"Kubectl Cert Renew Failed\""
|
||||
readonly DATA_APISERVER_CERT_OK="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE2 # \"ApiServer Cert Valid\""
|
||||
readonly DATA_LEAF_CERTS_RENEW_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE3 # \"K8S Leaf Certs Renew Failed\""
|
||||
readonly DATA_RENEW_FAIL_PODS="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE4 # \"Pods Cert Renew Failed\""
|
||||
readonly DATA_CERTMANAGER_CERTS_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE5 # \"Cert-manager Secrets Renew Failed\""
|
||||
readonly DATA_RENEW_FAIL_KUBECTL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE6 # \"Kubectl Cert Renew Failed\""
|
||||
|
||||
readonly DATA_MANUAL_CAS_OK="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe9 # \"Manual CA Certs Valid\""
|
||||
readonly DATA_MANUAL_CA_K8S_FP_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xec # \"K8S/Front-proxy Cert Expired\""
|
||||
readonly DATA_MANUAL_CA_ETCD_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xed # \"ETCD CA Cert Expired\""
|
||||
readonly DATA_MANUAL_CAS_BOTH_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xee # \"ETCD CA and K8S/Front-proxy Certs Expired\""
|
||||
readonly DATA_MANUAL_CAS_OK="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE7 # \"Manual CA Certs Valid\""
|
||||
readonly DATA_MANUAL_CA_K8S_FP_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE8 # \"K8S/Front-proxy Cert Expired\""
|
||||
readonly DATA_MANUAL_CA_ETCD_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xE9 # \"ETCD CA Cert Expired\""
|
||||
readonly DATA_MANUAL_CAS_BOTH_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xEA # \"ETCD CA and K8S/Front-proxy Certs Expired\""
|
||||
|
||||
readonly DATA_SLOCAL_CA_OK="0x04 0xF0 0x01 0x6f 0xff 0xff 0xea # \"System-local-ca Cert Valid\""
|
||||
readonly DATA_SLOCAL_CA_FAIL="0x04 0xF0 0x01 0x6f 0xff 0xff 0xeb # \"System-local-ca Cert Expired\""
|
||||
readonly DATA_SLOCAL_CA_OK="0x04 0x12 0xCC 0x63 0xCC 0x10 0xEB # \"System-local-ca Cert Valid\""
|
||||
readonly DATA_SLOCAL_CA_FAIL="0x04 0x12 0xCC 0x63 0xCC 0x10 0xEC # \"System-local-ca Cert Expired\""
|
||||
|
||||
readonly DATA_PLATFORM_CLOUDINIT_UPDATE_COMPLETE="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe4 # \"Platform Cloud-init Update Complete\""
|
||||
readonly DATA_PLATFORM_CLOUDINIT_UPDATE_FAILED="0x04 0xF0 0x01 0x6f 0xff 0xff 0xe5 # \"Platform Cloud-init Update Failed\""
|
||||
readonly DATA_PLATFORM_CLOUDINIT_UPDATE_COMPLETE="0x04 0x12 0xCC 0x63 0xCC 0x10 0xED # \"Platform Cloud-init Update Complete\""
|
||||
readonly DATA_PLATFORM_CLOUDINIT_UPDATE_FAILED="0x04 0x12 0xCC 0x63 0xCC 0x10 0xEE # \"Platform Cloud-init Update Failed\""
|
||||
|
||||
# IPMI SEL event format reference:
|
||||
# [EvM Revision] [Sensor Type] [Sensor Number] [Event Dir / Event Type Code]
|
||||
# [Event Data 1] [Event Data 2] [Event Data 3]
|
||||
#
|
||||
# Example: 0x04 0xF0 0x01 0x6f 0xff 0xff 0xe4
|
||||
# Example: 0x04 0x12 0xCC 0x63 0xCC 0x10 0xE0
|
||||
# 0x04 = EvM Revision (IPMI v2.0)
|
||||
# 0xF0 = Sensor Type (vendor-defined / OEM-specific)
|
||||
# 0x01 = Sensor Number (firmware-defined)
|
||||
# 0x6f = Event direction (vendor-specific encoding)
|
||||
# 0xff 0xff 0xe4 = Event Data bytes (3 bytes, OEM-specific payload)
|
||||
# 0x12 = Sensor Type (vendor-defined / OEM-specific)
|
||||
# 0xCC = Sensor Number (firmware-defined)
|
||||
# 0x63 = Event direction (vendor-specific encoding)
|
||||
# 0xCC 0x10 0xE0 = Event Data bytes (3 bytes, OEM-specific payload)
|
||||
#
|
||||
# For our usage:
|
||||
# - The third byte (Sensor Number) is set to 0x01, corresponding to a sensor type "Unknown".
|
||||
@@ -155,18 +155,6 @@ function send_ipmi_event {
|
||||
fi
|
||||
}
|
||||
|
||||
function verify_factory_install {
|
||||
log_info "Checking factory-install..."
|
||||
|
||||
if [ ! -f /var/lib/factory-install/stage/final ]; then
|
||||
send_ipmi_event "$EVENT_FACTORY_SETUP_FAILED"
|
||||
log_fatal "/var/lib/factory-install/stage/final does not exist. Ensure factory-install was successful."
|
||||
fi
|
||||
|
||||
send_ipmi_event "$EVENT_FACTORY_SETUP_COMPLETE"
|
||||
log_info "factory-install check successful."
|
||||
}
|
||||
|
||||
# The enroll-init reconfigure script runs during startup via cloud-init while
|
||||
# system services may not be settled. This timing can lead to intermittent errors
|
||||
# for early system commands. This function is used to mitigate these problems by
|
||||
@@ -639,11 +627,7 @@ while [[ "$#" -gt 0 ]]; do
|
||||
esac
|
||||
done
|
||||
|
||||
# This script can only be run if the factory install is complete, so we check that first.
|
||||
# It's important that we fail due to an invalid factory install before any other
|
||||
# type of failure, as that's the IPMI SEL event the system controller monitors first.
|
||||
# Main execution flow
|
||||
verify_factory_install
|
||||
|
||||
# Ensure all required arguments are provided
|
||||
if [ -z "$OAM_SUBNET" ] || [ -z "$OAM_GATEWAY_IP" ] || [ -z "$OAM_IP" ] || [ -z "$NEW_PASSWORD" ]; then
|
||||
|
||||
Reference in New Issue
Block a user