david/managed-container
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Remove development candidate files

Browse Source
This commit is contained in:
Lyncoln
2026-09-17 10:58:03 -04:00
parent b6c4fc848f
commit bfd2e1e7b2
4 changed files with 3 additions and 766 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.gitignore
+3
View File
@@ -18,3 +18,6 @@ __pycache__/
# Temporary files # Temporary files
*.tmp *.tmp
*.temp *.temp
# Development candidate files
*.new
ocf/loxthelion/managed-container.d/common.sh.named-rc.new
-209
View File
@@ -1,209 +0,0 @@
# Shared OCF scoring wrappers; application policy is supplied by the plugin.
# Promotion scores are scoped by Pacemaker to OCF_RESOURCE_INSTANCE and the
# executing node. Do not construct a legacy attribute name or cache locally:
# querying attrd on every observation also repairs a lost transient attribute.
mc_promotion_score_set()
{
local score="$1" current
local resource="${MC_PROMOTION_RESOURCE_INSTANCE:-}"
if [ -z "${resource}" ]; then
ocf_log warn "Cannot publish promotion score without OCF_RESOURCE_INSTANCE"
return 1
fi
case "${score}" in
1000|100|50|-INFINITY) ;;
*) return 1 ;;
esac
current="$(
timeout --kill-after=1s 2s crm_attribute \
--promotion="${resource}" --lifetime=reboot \
--query --quiet 2>/dev/null
)" || current=""
[ "${current}" = "${score}" ] && return 0
if ! timeout --kill-after=1s 2s crm_attribute \
--promotion="${resource}" --lifetime=reboot \
--update="${score}" >/dev/null 2>&1
then
ocf_log err "Unable to publish promotion score ${score} for ${resource}; prior score may remain"
return 1
fi
return 0
}
mc_promotion_score_from_monitor()
{
local monitor_rc="$1" score=-INFINITY
case "${monitor_rc}" in
"${OCF_RUNNING_PROMOTED}") score=1000 ;;
"${OCF_SUCCESS}")
score="$(mc_application_standby_score)" || score=-INFINITY
;;
esac
case "${score}" in
1000|100|50|-INFINITY) ;;
*) score=-INFINITY ;;
esac
mc_promotion_score_set "${score}"
}
# managed-containerd is authoritative for recurring monitor/status results.
# The legacy application monitor remains temporarily available as a shadow
# comparison only. Promotion scoring is still supplied by the existing
# application scoring hook during this migration stage.
mc_promotion_monitor()
{
local monitor_rc legacy_rc proxy_result proxy_rc cached_rc
proxy_result="$(mc_proxy_monitor)"
proxy_rc=$?
if [ "${proxy_rc}" -ne 0 ]; then
case "${proxy_rc}" in
"${MC_PROXY_RC_TRANSPORT}")
# Transport loss is a control-plane failure, not proof that
# the managed application itself has failed.
cached_rc="$(mc_proxy_cached_monitor 2>/dev/null)" || cached_rc=""
if [ -n "${cached_rc}" ]; then
monitor_rc="${cached_rc}"
case "${monitor_rc}" in
"${OCF_SUCCESS}")
# A standby with an unavailable control plane must
# never remain promotion-eligible during grace.
mc_promotion_score_set -INFINITY || true
;;
"${OCF_RUNNING_PROMOTED}")
# Preserve the established primary during a brief
# bounded transport outage.
;;
esac
ocf_log warn \
"managed-containerd unavailable; using recent cached authoritative rc=${monitor_rc} for bounded control-plane grace"
mc_application_monitor
legacy_rc=$?
mc_proxy_legacy_shadow_monitor \
"${monitor_rc}" \
"${legacy_rc}" || true
return "${monitor_rc}"
fi
mc_promotion_score_set -INFINITY || true
ocf_exit_reason \
"managed-containerd authoritative monitor transport failure and no valid control-plane grace observation exists"
return "${OCF_ERR_GENERIC}"
;;
*)
# Protocol/result/configuration/client failures are hard.
# Never conceal them behind cached authoritative health.
mc_proxy_monitor_cache_clear || true
mc_promotion_score_set -INFINITY || true
ocf_exit_reason \
"managed-containerd authoritative monitor hard failure; cached grace is not permitted"
return "${OCF_ERR_GENERIC}"
;;
esac
fi
monitor_rc="${proxy_result%%|*}"
case "${monitor_rc}" in
"${OCF_SUCCESS}"|\
"${OCF_ERR_GENERIC}"|\
"${OCF_NOT_RUNNING}"|\
"${OCF_RUNNING_PROMOTED}")
;;
*)
mc_promotion_score_set -INFINITY || true
ocf_exit_reason "managed-containerd returned invalid monitor rc ${monitor_rc}"
return "${OCF_ERR_GENERIC}"
;;
esac
ocf_log info "managed-containerd authoritative monitor: resource=${OCF_RESOURCE_INSTANCE:-unknown} proxy=${proxy_result}"
# Legacy PostgreSQL monitoring remains observational only.
mc_application_monitor
legacy_rc=$?
mc_proxy_legacy_shadow_monitor "${monitor_rc}" "${legacy_rc}" || true
mc_promotion_score_from_monitor "${monitor_rc}" || true
return "${monitor_rc}"
}
# Lifecycle functions retain their original checks, commands and return codes.
# Withdraw the previous score before any transition (including validation
# failure), then reassess only after success. Their internal monitor calls do
# not publish scores, so they cannot restore a positive score mid-transition.
mc_promotion_action()
{
local action="$1" action_rc monitor_rc
case "${action}" in
start|stop|promote|demote|rejoin) ;;
*) return "${OCF_ERR_UNIMPLEMENTED}" ;;
esac
# Any lifecycle mutation invalidates the previous daemon-observed role.
# A fresh daemon monitor must establish the next cache entry.
mc_proxy_monitor_cache_clear
mc_promotion_score_set -INFINITY || true
if [ "${action}" != stop ]; then
mc_promotion_validate
action_rc=$?
[ "${action_rc}" -eq "${OCF_SUCCESS}" ] || return "${action_rc}"
fi
mc_application_action "${action}"
action_rc=$?
if [ "${action_rc}" -eq "${OCF_SUCCESS}" ] && [ "${action}" != stop ]; then
mc_application_monitor
monitor_rc=$?
mc_promotion_score_from_monitor "${monitor_rc}" || true
fi
return "${action_rc}"
}
mc_promotion_validate()
{
local utility
if ! [[ "${OCF_RESKEY_proxy_failure_grace_seconds:-15}" =~ ^(0|[1-9][0-9]{0,2})$ ]]; then
ocf_exit_reason "proxy_failure_grace_seconds must be an integer from 0 to 999 seconds"
return "${OCF_ERR_CONFIGURED}"
fi
if [ "${OCF_RESKEY_proxy_failure_grace_seconds:-15}" -gt 300 ]; then
ocf_exit_reason "proxy_failure_grace_seconds must be between 0 and 300 seconds"
return "${OCF_ERR_CONFIGURED}"
fi
if ! [[ "${OCF_RESKEY_disconnected_max_age:-300}" =~ ^(0|[1-9][0-9]{0,4})$ ]]; then
ocf_exit_reason "disconnected_max_age must be an integer from 0 to 99999 seconds"
return "${OCF_ERR_CONFIGURED}"
fi
for utility in crm_attribute timeout python3; do
if ! command -v "${utility}" >/dev/null 2>&1; then
ocf_exit_reason "Required promotion scoring command '${utility}' is not installed"
return "${OCF_ERR_INSTALLED}"
fi
done
mc_application_validate
}
ocf/loxthelion/managed-container.d/proxy.sh.named-rc.new
-464
View File
@@ -1,464 +0,0 @@
# Generic managed-containerd client helpers.
# Shadow calls are observational only and must never affect OCF action results.
MC_PROXY_SOCKET_DEFAULT="/run/user/1000/managed-containerd/managed-containerd.sock"
mc_proxy_shadow_monitor()
{
local authoritative_rc="$1"
local socket_path="${MC_PROXY_SOCKET:-${MC_PROXY_SOCKET_DEFAULT}}"
local resource="${OCF_RESOURCE_INSTANCE:-}"
local response
resource="${resource%%:*}"
[ -n "${resource}" ] || return 0
[ -S "${socket_path}" ] || return 0
command -v python3 >/dev/null 2>&1 || return 0
response="$(
timeout --kill-after=1s 3s python3 - "${socket_path}" "${resource}" "${authoritative_rc}" <<'PY'
import json
import socket
import sys
path, resource, authoritative_raw = sys.argv[1], sys.argv[2], sys.argv[3]
authoritative_rc = int(authoritative_raw)
request = {"action": "monitor", "resource": resource}
s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
s.settimeout(2.0)
s.connect(path)
s.sendall(json.dumps(request).encode("utf-8") + b"\n")
raw = b""
while b"\n" not in raw:
chunk = s.recv(65536)
if not chunk:
break
raw += chunk
s.close()
obj = json.loads(raw.decode("utf-8").strip())
if not obj.get("ok"):
print("ERROR|" + str(obj.get("error", "unknown proxy error")))
raise SystemExit(0)
result = obj.get("result") or {}
ha = result.get("ha") or {}
proxy_rc = ha.get("ocf_rc")
if proxy_rc is None:
comparison = "UNCLASSIFIED"
else:
comparison = "MATCH" if int(proxy_rc) == authoritative_rc else "MISMATCH"
receiver = result.get("wal_receiver") or {}
print("OK|{}|{}|{}|{}|{}|{}|{}|{}|{}|{}".format(
comparison,
"null" if proxy_rc is None else proxy_rc,
ha.get("health", "unknown"),
ha.get("ha_role", "unknown"),
ha.get("state", "unknown"),
str(ha.get("promotion_eligible", "unknown")).lower(),
result.get("role", "unknown"),
str(result.get("streaming", "unknown")).lower(),
receiver.get("slot_name") or "",
receiver.get("sender_host") or "",
))
PY
)" || response="TRANSPORT_ERROR"
case "${response}" in
OK\|MATCH\|*)
ocf_log info \
"managed-containerd shadow monitor: resource=${resource} authoritative_rc=${authoritative_rc} proxy=${response}"
;;
OK\|UNCLASSIFIED\|*)
ocf_log warn \
"managed-containerd shadow monitor unclassified: resource=${resource} authoritative_rc=${authoritative_rc} proxy=${response}"
;;
OK\|MISMATCH\|*)
ocf_log err \
"managed-containerd shadow monitor MISMATCH: resource=${resource} authoritative_rc=${authoritative_rc} proxy=${response}"
;;
ERROR\|*)
ocf_log warn \
"managed-containerd shadow monitor failed: resource=${resource} authoritative_rc=${authoritative_rc} proxy=${response}"
;;
*)
ocf_log warn \
"managed-containerd shadow monitor transport failure: resource=${resource} authoritative_rc=${authoritative_rc}"
;;
esac
return 0
}
# Query managed-containerd for the authoritative monitor result.
#
# stdout:
# <OCF rc>|<normalized diagnostic payload>
#
# return:
# 0 valid classified daemon result
# MC_PROXY_RC_TRANSPORT daemon transport/control-plane failure; bounded cache grace allowed
# MC_PROXY_RC_HARD_FAILURE configuration/protocol/result failure; fail closed, no cache grace
mc_proxy_monitor()
{
local socket_path="${MC_PROXY_SOCKET:-${MC_PROXY_SOCKET_DEFAULT}}"
local resource="${OCF_RESOURCE_INSTANCE:-}"
local response rc
resource="${resource%%:*}"
if [ -z "${resource}" ]; then
ocf_log err \
"managed-containerd monitor requires OCF_RESOURCE_INSTANCE"
return "${MC_PROXY_RC_HARD_FAILURE}"
fi
if [ ! -S "${socket_path}" ]; then
ocf_log warn \
"managed-containerd socket is unavailable: ${socket_path}"
return "${MC_PROXY_RC_TRANSPORT}"
fi
if ! command -v python3 >/dev/null 2>&1; then
ocf_log err \
"python3 is required for managed-containerd monitor transport"
return "${MC_PROXY_RC_HARD_FAILURE}"
fi
response="$(
timeout --kill-after=1s 3s \
python3 - \
"${socket_path}" \
"${resource}" \
"${MC_PROXY_RC_TRANSPORT}" \
"${MC_PROXY_RC_HARD_FAILURE}" <<'PY'
import json
import socket
import sys
path, resource = sys.argv[1], sys.argv[2]
TRANSPORT_FAILURE = int(sys.argv[3])
PROTOCOL_FAILURE = int(sys.argv[4])
request = {
"action": "monitor",
"resource": resource,
}
try:
s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
s.settimeout(2.0)
try:
s.connect(path)
s.sendall(json.dumps(request).encode("utf-8") + b"\n")
raw = b""
while b"\n" not in raw:
chunk = s.recv(65536)
if not chunk:
break
raw += chunk
finally:
s.close()
except (socket.timeout, ConnectionError, OSError):
# Local daemon/control-plane reachability failure.
# Recent authoritative state may be used for bounded grace.
raise SystemExit(TRANSPORT_FAILURE)
# Transport succeeded. Invalid daemon output is therefore a
# protocol/result failure and must never be hidden by cached health.
if not raw:
raise SystemExit(PROTOCOL_FAILURE)
try:
obj = json.loads(raw.decode("utf-8").strip())
except (UnicodeDecodeError, json.JSONDecodeError):
raise SystemExit(PROTOCOL_FAILURE)
if not isinstance(obj, dict):
raise SystemExit(PROTOCOL_FAILURE)
if not obj.get("ok"):
raise SystemExit(PROTOCOL_FAILURE)
result = obj.get("result")
if not isinstance(result, dict):
raise SystemExit(PROTOCOL_FAILURE)
ha = result.get("ha")
if not isinstance(ha, dict):
raise SystemExit(PROTOCOL_FAILURE)
rc = ha.get("ocf_rc")
try:
rc = int(rc)
except (TypeError, ValueError):
raise SystemExit(PROTOCOL_FAILURE)
if rc not in (0, 1, 7, 8):
raise SystemExit(PROTOCOL_FAILURE)
receiver = result.get("wal_receiver") or {}
if not isinstance(receiver, dict):
raise SystemExit(PROTOCOL_FAILURE)
payload = "|".join([
str(rc),
str(ha.get("health", "unknown")),
str(ha.get("ha_role", "unknown")),
str(ha.get("state", "unknown")),
str(ha.get("promotion_eligible", "unknown")).lower(),
str(result.get("role", "unknown")),
str(result.get("streaming", "unknown")).lower(),
str(receiver.get("slot_name") or ""),
str(receiver.get("sender_host") or ""),
])
print(payload)
PY
)"
rc=$?
case "${rc}" in
0)
;;
"${MC_PROXY_RC_TRANSPORT}"|124|137)
# MC_PROXY_RC_TRANSPORT = classified Python transport failure.
# 124/137 = timeout(1) expiry / forced termination.
ocf_log warn \
"managed-containerd authoritative monitor transport failure"
return "${MC_PROXY_RC_TRANSPORT}"
;;
"${MC_PROXY_RC_HARD_FAILURE}")
ocf_log err \
"managed-containerd authoritative monitor protocol or result failure"
return "${MC_PROXY_RC_HARD_FAILURE}"
;;
*)
ocf_log err \
"managed-containerd authoritative monitor client failure rc=${rc}"
return "${MC_PROXY_RC_HARD_FAILURE}"
;;
esac
if [ -z "${response}" ]; then
ocf_log err \
"managed-containerd authoritative monitor returned an empty result"
return "${MC_PROXY_RC_HARD_FAILURE}"
fi
rc="${response%%|*}"
# Cache only healthy daemon-authoritative role observations. An
# application-level failure automatically invalidates any previous cache.
mc_proxy_monitor_cache_write "${rc}" || {
ocf_log warn \
"Unable to update managed-containerd authoritative monitor cache"
}
printf '%s\n' "${response}"
return 0
}
# Compare the legacy application monitor against a daemon-authoritative rc.
# The legacy result is diagnostic only and can never change the returned
# Pacemaker monitor status.
mc_proxy_legacy_shadow_monitor()
{
local authoritative_rc="$1"
local legacy_rc="$2"
local resource="${OCF_RESOURCE_INSTANCE:-}"
resource="${resource%%:*}"
if [ "${legacy_rc}" -eq "${authoritative_rc}" ]; then
ocf_log info \
"legacy application shadow monitor: resource=${resource} authoritative_rc=${authoritative_rc} legacy_rc=${legacy_rc} comparison=MATCH"
else
ocf_log err \
"legacy application shadow monitor MISMATCH: resource=${resource} authoritative_rc=${authoritative_rc} legacy_rc=${legacy_rc}"
fi
return 0
}
# Generic short-lived cache of the last healthy daemon-authoritative OCF role.
# This is control-plane state only; it contains no application-specific data.
#
# Use HA_RSCTMP when available because it is the standard resource-agent
# runtime location and is preferable to inventing a persistent filesystem path.
MC_PROXY_STATE_DIR_DEFAULT="${HA_RSCTMP:-/run/resource-agents}/managed-container"
mc_proxy_state_file()
{
local resource safe_resource
resource="${OCF_RESOURCE_INSTANCE:-}"
resource="${resource%%:*}"
[ -n "${resource}" ] || return 1
safe_resource="$(
printf '%s' "${resource}" |
tr -c 'A-Za-z0-9_.-' '_'
)"
printf '%s/%s.monitor\n' \
"${MC_PROXY_STATE_DIR:-${MC_PROXY_STATE_DIR_DEFAULT}}" \
"${safe_resource}"
}
mc_proxy_monitor_cache_clear()
{
local state_file
state_file="$(mc_proxy_state_file)" || return 0
rm -f "${state_file}" 2>/dev/null || true
}
mc_proxy_monitor_cache_write()
{
local rc="$1"
local state_file state_dir tmp now
# Only healthy authoritative role observations are eligible for
# temporary control-plane grace.
case "${rc}" in
"${OCF_SUCCESS}"|"${OCF_RUNNING_PROMOTED}")
;;
*)
mc_proxy_monitor_cache_clear
return 0
;;
esac
state_file="$(mc_proxy_state_file)" || return 1
state_dir="${state_file%/*}"
mkdir -p -m 700 "${state_dir}" || return 1
now="$(date +%s)"
tmp="${state_file}.$$"
umask 077
printf '%s|%s|0
' "${now}" "${rc}" > "${tmp}" || {
rm -f "${tmp}" 2>/dev/null || true
return 1
}
chmod 600 "${tmp}" 2>/dev/null || true
mv -f "${tmp}" "${state_file}"
}
# Print a cached healthy OCF rc when its observation is still inside the
# configured daemon control-plane failure grace.
mc_proxy_cached_monitor()
{
local state_file observed rc failure_started
local now failure_age observation_age grace max_observation_age
local tmp
grace="${OCF_RESKEY_proxy_failure_grace_seconds:-15}"
case "${grace}" in
""|*[!0-9]*)
return 1
;;
esac
[ "${grace}" -gt 0 ] || return 1
state_file="$(mc_proxy_state_file)" || return 1
[ -f "${state_file}" ] || return 1
IFS='|' read -r observed rc failure_started < "${state_file}" || {
mc_proxy_monitor_cache_clear
return 1
}
case "${observed}" in
""|*[!0-9]*)
mc_proxy_monitor_cache_clear
return 1
;;
esac
case "${failure_started}" in
""|*[!0-9]*)
mc_proxy_monitor_cache_clear
return 1
;;
esac
case "${rc}" in
"${OCF_SUCCESS}"|"${OCF_RUNNING_PROMOTED}")
;;
*)
mc_proxy_monitor_cache_clear
return 1
;;
esac
now="$(date +%s)"
observation_age=$((now - observed))
# A cached role must come from a reasonably recent successful daemon
# observation. This protects against resurrecting stale runtime state.
max_observation_age=$((grace + 60))
if [ "${observation_age}" -lt 0 ] ||
[ "${observation_age}" -gt "${max_observation_age}" ]; then
mc_proxy_monitor_cache_clear
return 1
fi
# First failed daemon observation starts the bounded grace period.
if [ "${failure_started}" -eq 0 ]; then
failure_started="${now}"
tmp="${state_file}.$$"
umask 077
printf '%s|%s|%s
' "${observed}" "${rc}" "${failure_started}" > "${tmp}" || {
rm -f "${tmp}" 2>/dev/null || true
return 1
}
chmod 600 "${tmp}" 2>/dev/null || true
mv -f "${tmp}" "${state_file}" || return 1
fi
failure_age=$((now - failure_started))
if [ "${failure_age}" -lt 0 ] ||
[ "${failure_age}" -ge "${grace}" ]; then
mc_proxy_monitor_cache_clear
return 1
fi
printf '%s
' "${rc}"
return 0
}
ocf/loxthelion/managed-container.named-rc.new
-93
View File
@@ -1,93 +0,0 @@
#!/bin/bash
# OCF Resource Agent:
# loxthelion:managed-container
#
# v0.4
# Container lifecycle: systemd user Quadlet
# Application profile: PostgreSQL
#
# Version 0.4 adds explicit PostgreSQL demotion scaffolding.
# Healthy primaries report OCF_RUNNING_PROMOTED while healthy standbys
# report OCF_SUCCESS.
: "${OCF_ROOT:=/usr/lib/ocf}"
OCF_FUNCTIONS_DIR="${OCF_ROOT}/lib/heartbeat"
# Preserve explicit resource identity before ocf-shellfuncs supplies its
# standalone-invocation default; never publish a score for that dummy resource.
MC_PROMOTION_RESOURCE_INSTANCE="${OCF_RESOURCE_INSTANCE:-}"
. "${OCF_FUNCTIONS_DIR}/ocf-shellfuncs"
# Pacemaker uses "Promoted" terminology, while many OCF shell libraries
# still expose the historical OCF_RUNNING_MASTER constant for return code 8.
# Normalize both forms here so the agent remains portable.
: "${OCF_RUNNING_PROMOTED:=${OCF_RUNNING_MASTER:-8}}"
# Internal managed-containerd client return classes.
# Private RA implementation statuses, not OCF action return codes.
readonly MC_PROXY_RC_TRANSPORT=10
readonly MC_PROXY_RC_HARD_FAILURE=11
# Resolve modules relative to this entrypoint, independent of Pacemaker's cwd.
MC_MODULE_DIR="$(CDPATH= cd -- "$(command dirname -- "${BASH_SOURCE[0]}")" && pwd)/managed-container.d"
for MC_MODULE in common container engine; do
if ! . "${MC_MODULE_DIR}/${MC_MODULE}.sh"; then
ocf_exit_reason "Unable to load resource-agent module ${MC_MODULE_DIR}/${MC_MODULE}.sh"
exit "${OCF_ERR_CONFIGURED}"
fi
done
unset MC_MODULE
mc_engine_load || exit $?
case "$1" in
meta-data)
mc_engine_metadata
exit $?
;;
validate-all)
mc_promotion_validate
exit $?
;;
monitor|status)
mc_promotion_monitor
exit $?
;;
start)
mc_promotion_action start
exit $?
;;
stop)
mc_promotion_action stop
exit $?
;;
promote)
mc_promotion_action promote
exit $?
;;
demote)
mc_promotion_action demote
exit $?
;;
notify)
mc_application_notify
exit $?
;;
rejoin)
mc_promotion_action rejoin
exit $?
;;
*)
echo "Usage: $0 {start|stop|promote|demote|notify|rejoin|monitor|status|validate-all|meta-data}"
exit "${OCF_ERR_UNIMPLEMENTED}"
;;
esac