verify all heavynodes and all searchnodes are at compatible ES version before attempting an elasticsearch upgrade

This commit is contained in:
reyesj2
2026-05-19 22:27:31 -05:00
parent 74b50f6009
commit 6c8997b28a
+134 -5
View File
@@ -971,6 +971,9 @@ verify_es_version_compatibility() {
local is_active_intermediate_upgrade=1
# supported upgrade paths for SO-ES versions
declare -A es_upgrade_map=(
["8.18.4"]="8.18.6 8.18.8 9.0.8"
["8.18.6"]="8.18.8 9.0.8"
["8.18.8"]="9.0.8"
["9.0.8"]="9.3.3"
)
@@ -994,6 +997,116 @@ verify_es_version_compatibility() {
exit 160
fi
compatible_es_versions="$target_es_version"
for current_version in "${!es_upgrade_map[@]}"; do
# shellcheck disable=SC2076
if [[ " ${es_upgrade_map[$current_version]} " =~ " $target_es_version " ]]; then
compatible_es_versions+=" $current_version"
fi
done
# Check if the given ES version can directly upgrade to the target ES version. Used to assist with catching lagging nodes during the upgrade process
es_version_can_upgrade_to_target() {
local current_version="$1"
# shellcheck disable=SC2076
if [[ -n "$current_version" && " $compatible_es_versions " =~ " $current_version " ]]; then
return 0
fi
return 1
}
# Gather Elasticsearch cluster version info and verify that each node in the cluster is running a version compatible with the target ES version.
verify_searchnodes_es_target_compatibility() {
local retries=20
local retry_count=0
local delay=180
SEARCHNODE_ES_VERSIONS=""
while [[ $retry_count -lt $retries ]]; do
SEARCHNODE_ES_VERSIONS=$(so-elasticsearch-query _nodes/_all/version --retry 5 --retry-delay 10 --fail 2>&1)
local exit_status=$?
if [[ $exit_status -ne 0 ]]; then
echo "Failed to retrieve Elasticsearch versions from searchnodes... Retrying in $delay seconds. Attempt $((retry_count + 1)) of $retries."
((retry_count++))
sleep $delay
continue
fi
local all_searchnodes_compatible=true
while IFS=$'\t' read -r node current_version; do
[[ -z "$node" ]] && continue
if ! es_version_can_upgrade_to_target "$current_version"; then
echo "Searchnode $node is running Elasticsearch $current_version, which is not directly upgradable to Elasticsearch $target_es_version."
all_searchnodes_compatible=false
fi
done < <(echo "$SEARCHNODE_ES_VERSIONS" | jq -r '.nodes | to_entries[] | [.value.name, .value.version] | @tsv')
if [[ "$all_searchnodes_compatible" == true ]]; then
echo "All Searchnodes are upgradable to Elasticsearch $target_es_version."
return 0
fi
echo "One or more Searchnodes cannot upgrade directly to Elasticsearch $target_es_version. Rechecking in $delay seconds. Attempt $((retry_count + 1)) of $retries."
((retry_count++))
sleep $delay
done
return 1
}
# Gather heavynode version info and verify that each node is running a version compatible with the target ES version.
verify_heavynodes_es_target_compatibility() {
if ! salt-key -l accepted | grep -q 'heavynode$'; then
echo "No heavynodes detected. Skipping heavynode Elasticsearch version compatibility check."
return 0
fi
echo -e "\nOne or more heavynodes detected. Verifying each is running an Elasticsearch version that is compatible with $target_es_version."
local retries=20
local retry_count=0
local delay=180
HEAVYNODE_ES_VERSIONS=""
while [[ $retry_count -lt $retries ]]; do
HEAVYNODE_ES_VERSIONS=$(salt -C 'G@role:so-heavynode' cmd.run 'so-elasticsearch-query / --retry 5 --retry-delay 10 | jq -r ".version.number"' shell=/bin/bash --out=json 2> /dev/null)
local exit_status=$?
if [[ $exit_status -ne 0 ]]; then
echo "Failed to retrieve Elasticsearch version from one or more heavynodes... Retrying in $delay seconds. Attempt $((retry_count + 1)) of $retries."
((retry_count++))
sleep $delay
continue
fi
local all_heavynodes_compatible=true
while IFS=$'\t' read -r node current_version; do
[[ -z "$node" ]] && continue
if ! es_version_can_upgrade_to_target "$current_version"; then
echo "Heavynode $node is running Elasticsearch $current_version, which is not directly upgradable to Elasticsearch $target_es_version."
all_heavynodes_compatible=false
fi
done < <(echo "$HEAVYNODE_ES_VERSIONS" | jq -r 'to_entries[] | [.key, .value] | @tsv')
if [[ "$all_heavynodes_compatible" == true ]]; then
echo -e "\nAll heavynodes can upgrade to Elasticsearch $target_es_version."
return 0
fi
echo "One or more heavynodes cannot upgrade directly to Elasticsearch $target_es_version. Rechecking in $delay seconds. Attempt $((retry_count + 1)) of $retries."
((retry_count++))
sleep $delay
done
return 1
}
if [[ ! -f "$es_verification_script" ]]; then
create_intermediate_upgrade_verification_script "$es_verification_script"
fi
for statefile in "${es_required_version_statefile_base}"-*; do
[[ -f $statefile ]] || continue
@@ -1012,10 +1125,6 @@ verify_es_version_compatibility() {
continue
fi
if [[ ! -f "$es_verification_script" ]]; then
create_intermediate_upgrade_verification_script "$es_verification_script"
fi
echo -e "\n##############################################################################################################################\n"
echo "A previously required intermediate Elasticsearch upgrade was detected. Verifying that all Searchnodes/Heavynodes have successfully upgraded Elasticsearch to $es_required_version_statefile_value before proceeding with soup to avoid potential data loss! This command can take up to an hour to complete."
if ! timeout --foreground 4000 bash "$es_verification_script" "$es_required_version_statefile_value" "$statefile"; then
@@ -1037,6 +1146,26 @@ verify_es_version_compatibility() {
# shellcheck disable=SC2076 # Do not want a regex here eg usage " 8.18.8 9.0.8 " =~ " 9.0.8 "
if [[ " ${es_upgrade_map[$es_version]} " =~ " $target_es_version " || "$es_version" == "$target_es_version" ]]; then
if ! verify_searchnodes_es_target_compatibility || ! verify_heavynodes_es_target_compatibility; then
echo -e "\n!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n"
echo "One or more Searchnode(s)/Heavynode(s) cannot upgrade directly to Elasticsearch $target_es_version. This can happen with soups that include Elasticsearch upgrades being run in quick succession. Typically, this will resolve itself as the grid synchronizes. Please allow time for all Searchnodes/Heavynodes to have upgraded Elasticsearch to a compatible version with $target_es_version before running soup again to avoid potential data loss!"
if [[ -n "$HEAVYNODE_ES_VERSIONS" ]]; then
echo "Current heavynode Elasticsearch versions:"
echo "$HEAVYNODE_ES_VERSIONS" | jq '.'
fi
if [[ -n "$SEARCHNODE_ES_VERSIONS" ]]; then
echo "Current searchnode Elasticsearch versions:"
echo "$SEARCHNODE_ES_VERSIONS" | jq '.nodes | to_entries | map({(.value.name): .value.version}) | sort | add'
fi
echo -e "\n!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n"
exit 161
fi
# supported upgrade
return 0
else
@@ -1394,7 +1523,7 @@ main() {
echo "Verifying we have the latest soup script."
verify_latest_update_script
echo "Verifying Elasticsearch version compatibility before upgrading."
echo "Verifying Elasticsearch version compatibility across the grid before upgrading."
verify_es_version_compatibility
echo "Let's see if we need to update Security Onion."