Author SHA1 Message Date
Lyncoln 3d1714d995 Remove legacy monitor shadow execution 2026-09-18 15:13:13 -04:00
Lyncoln cbb865729a Add daemon-selected container monitor delegation 2026-09-18 14:09:46 -04:00
6 changed files with 418 additions and 17 deletions
+71 -17
View File
@@ -54,12 +54,12 @@ mc_promotion_score_from_monitor()
}
# 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.
# The packaged application monitor is used only when managed-containerd
# explicitly delegates monitoring to the container backend. Promotion scoring
# remains supplied by the existing application scoring hook.
mc_promotion_monitor()
{
local monitor_rc legacy_rc proxy_result proxy_rc cached_rc
local monitor_rc proxy_result proxy_rc cached_rc
proxy_result="$(mc_proxy_monitor)"
proxy_rc=$?
@@ -91,13 +91,6 @@ mc_promotion_monitor()
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 "${monitor_rc}"
fi
@@ -119,6 +112,70 @@ mc_promotion_monitor()
esac
fi
# A successful managed-containerd response may explicitly delegate
# monitoring to the RA's already-loaded packaged container/runtime path.
#
# The daemon remains authoritative for backend selection: this path is
# reachable only after managed-containerd successfully requested it.
case "${proxy_result}" in
delegate\|*)
local delegate_tag delegate_backend delegate_runtime delegate_extra
local configured_runtime
IFS='|' read -r delegate_tag delegate_backend delegate_runtime delegate_extra <<EOF
${proxy_result}
EOF
if [ "${delegate_tag}" != delegate ] ||
[ "${delegate_backend}" != container ] ||
[ -z "${delegate_runtime}" ] ||
[ -n "${delegate_extra}" ]; then
mc_proxy_monitor_cache_clear || true
mc_promotion_score_set -INFINITY || true
ocf_exit_reason "managed-containerd returned invalid monitor delegation: ${proxy_result}"
return "${OCF_ERR_GENERIC}"
fi
configured_runtime="${OCF_RESKEY_container_runtime:-podman}"
if [ "${delegate_runtime}" != "${configured_runtime}" ]; then
mc_proxy_monitor_cache_clear || true
mc_promotion_score_set -INFINITY || true
ocf_exit_reason "managed-containerd delegated runtime ${delegate_runtime} does not match loaded RA runtime ${configured_runtime}"
return "${OCF_ERR_GENERIC}"
fi
ocf_log info "managed-containerd delegated authoritative monitor: resource=${OCF_RESOURCE_INSTANCE:-unknown} backend=${delegate_backend} runtime=${delegate_runtime}"
mc_application_monitor
monitor_rc=$?
case "${monitor_rc}" in
"${OCF_SUCCESS}"| "${OCF_ERR_GENERIC}"| "${OCF_NOT_RUNNING}"| "${OCF_RUNNING_PROMOTED}")
;;
*)
mc_proxy_monitor_cache_clear || true
mc_promotion_score_set -INFINITY || true
ocf_exit_reason "delegated application monitor returned invalid rc ${monitor_rc}"
return "${OCF_ERR_GENERIC}"
;;
esac
# The delegated application result is now the authoritative
# observation. Cache only healthy 0/8 results using the same
# control-plane grace mechanism as daemon-native monitoring.
mc_proxy_monitor_cache_write "${monitor_rc}" || {
ocf_log warn "Unable to update delegated authoritative monitor cache"
}
ocf_log info "managed-containerd delegated monitor result: resource=${OCF_RESOURCE_INSTANCE:-unknown} rc=${monitor_rc} backend=${delegate_backend} runtime=${delegate_runtime}"
mc_promotion_score_from_monitor "${monitor_rc}" || true
return "${monitor_rc}"
;;
esac
monitor_rc="${proxy_result%%|*}"
case "${monitor_rc}" in
@@ -136,12 +193,9 @@ mc_promotion_monitor()
ocf_log info "managed-containerd authoritative monitor: resource=${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
# The daemon-native monitor is authoritative. The packaged application
# monitor is invoked only when managed-containerd explicitly delegates to
# the container backend.
mc_promotion_score_from_monitor "${monitor_rc}" || true
return "${monitor_rc}"
}
@@ -88,6 +88,16 @@ mc_proxy_monitor()
return "${MC_PROXY_RC_HARD_FAILURE}"
fi
# Delegated monitor execution is a valid control-plane response, but it is
# not itself an OCF observation and must never be written to the healthy
# role cache. The caller validates and executes the packaged runtime path.
case "${response}" in
delegate\|*)
printf '%s\n' "${response}"
return 0
;;
esac
rc="${response%%|*}"
# Cache only healthy daemon-authoritative role observations. An
@@ -117,6 +117,22 @@ def format_monitor_result(result):
if not isinstance(result, dict):
raise ProtocolFailure
if result.get("delegated") is True:
backend = result.get("backend")
runtime = result.get("container_runtime")
if backend != "container":
raise ProtocolFailure
if runtime not in ("podman", "docker"):
raise ProtocolFailure
return "|".join([
"delegate",
backend,
runtime,
])
ha = result.get("ha")
if not isinstance(ha, dict):
+174
View File
@@ -0,0 +1,174 @@
#!/usr/bin/env bash
set -u
OCF_SUCCESS=0
OCF_ERR_GENERIC=1
OCF_NOT_RUNNING=7
OCF_RUNNING_PROMOTED=8
MC_PROXY_RC_TRANSPORT=10
MC_PROXY_RC_HARD_FAILURE=11
OCF_RESOURCE_INSTANCE=postgresql-ha
OCF_RESKEY_container_runtime=podman
logs=()
exit_reason=""
score=""
cache_rc=""
application_rc=0
application_calls=0
proxy_result_value="delegate|container|podman"
# Load the real function under test first. Test doubles defined below then
# override its external dependencies, including promotion-score helpers.
source ocf/loxthelion/managed-container.d/common.sh
ocf_log()
{
logs+=("$*")
}
ocf_exit_reason()
{
exit_reason="$*"
}
mc_proxy_monitor()
{
printf '%s\n' "${proxy_result_value}"
return 0
}
mc_proxy_cached_monitor()
{
return 1
}
mc_proxy_monitor_cache_clear()
{
cache_rc="cleared"
return 0
}
mc_proxy_monitor_cache_write()
{
cache_rc="$1"
return 0
}
mc_proxy_legacy_shadow_monitor()
{
echo "ERROR: shadow monitor must not execute for delegation" >&2
return 99
}
mc_application_monitor()
{
application_calls=$((application_calls + 1))
return "${application_rc}"
}
mc_promotion_score_set()
{
score="$1"
return 0
}
mc_promotion_score_from_monitor()
{
case "$1" in
0|8)
score="healthy:$1"
;;
*)
score="-INFINITY"
;;
esac
return 0
}
echo "===== PODMAN DELEGATION SUCCESS ====="
application_rc=0
proxy_result_value="delegate|container|podman"
cache_rc=""
score=""
application_calls=0
mc_promotion_monitor
rc=$?
[ "$rc" -eq 0 ]
[ "$application_calls" -eq 1 ]
[ "$cache_rc" = 0 ]
[ "$score" = "healthy:0" ]
echo "podman_success accepted"
echo
echo "===== PROMOTED DELEGATION ====="
application_rc=8
proxy_result_value="delegate|container|podman"
cache_rc=""
score=""
application_calls=0
mc_promotion_monitor
rc=$?
[ "$rc" -eq 8 ]
[ "$application_calls" -eq 1 ]
[ "$cache_rc" = 8 ]
[ "$score" = "healthy:8" ]
echo "podman_promoted accepted"
echo
echo "===== RUNTIME MISMATCH ====="
application_rc=0
proxy_result_value="delegate|container|docker"
cache_rc=""
score=""
application_calls=0
mc_promotion_monitor
rc=$?
[ "$rc" -eq 1 ]
[ "$application_calls" -eq 0 ]
[ "$cache_rc" = "cleared" ]
[ "$score" = "-INFINITY" ]
echo "runtime_mismatch rejected"
echo
echo "===== MALFORMED DELEGATION ====="
proxy_result_value="delegate|container|podman|extra"
cache_rc=""
score=""
application_calls=0
mc_promotion_monitor
rc=$?
[ "$rc" -eq 1 ]
[ "$application_calls" -eq 0 ]
[ "$cache_rc" = "cleared" ]
[ "$score" = "-INFINITY" ]
echo "malformed_delegation rejected"
echo
echo "MONITOR_DELEGATION_SHELL_TEST_PASS=1"
+139
View File
@@ -0,0 +1,139 @@
#!/usr/bin/env python3
import importlib.util
from pathlib import Path
spec = importlib.util.spec_from_file_location(
"proxy_client",
Path(
"ocf/loxthelion/"
"managed-container.d/"
"proxy_client.py"
),
)
proxy = importlib.util.module_from_spec(spec)
spec.loader.exec_module(proxy)
print("===== NORMAL API MONITOR RESULT =====")
normal = {
"ha": {
"ocf_rc": 0,
"health": "healthy",
"ha_role": "unpromoted",
"state": "streaming",
"promotion_eligible": True,
},
"role": "standby",
"streaming": True,
"wal_receiver": {
"slot_name": "ha_pg_node1",
"sender_host": "192.0.2.10",
},
}
formatted = proxy.format_monitor_result(normal)
expected = (
"0|healthy|unpromoted|streaming|true|"
"standby|true|ha_pg_node1|192.0.2.10"
)
assert formatted == expected, formatted
print("api_monitor_format accepted")
print()
print("===== PODMAN DELEGATION =====")
delegated = {
"delegated": True,
"backend": "container",
"container_runtime": "podman",
}
formatted = proxy.format_monitor_result(delegated)
assert formatted == "delegate|container|podman", formatted
print("podman_delegation accepted")
print()
print("===== DOCKER DELEGATION =====")
delegated = {
"delegated": True,
"backend": "container",
"container_runtime": "docker",
}
formatted = proxy.format_monitor_result(delegated)
assert formatted == "delegate|container|docker", formatted
print("docker_delegation accepted")
print()
print("===== INVALID DELEGATED BACKEND =====")
bad = {
"delegated": True,
"backend": "api",
"container_runtime": "podman",
}
try:
proxy.format_monitor_result(bad)
except proxy.ProtocolFailure:
print("invalid_backend rejected")
else:
raise AssertionError(
"delegation with non-container backend was accepted"
)
print()
print("===== INVALID DELEGATED RUNTIME =====")
bad = {
"delegated": True,
"backend": "container",
"container_runtime": "lxc",
}
try:
proxy.format_monitor_result(bad)
except proxy.ProtocolFailure:
print("invalid_runtime rejected")
else:
raise AssertionError(
"delegation with invalid runtime was accepted"
)
print()
print("===== MALFORMED DELEGATION =====")
bad = {
"delegated": True,
"backend": "container",
}
try:
proxy.format_monitor_result(bad)
except proxy.ProtocolFailure:
print("missing_runtime rejected")
else:
raise AssertionError(
"delegation without runtime was accepted"
)
print()
print("PROXY_MONITOR_DELEGATION_TEST_PASS=1")
+8
View File
@@ -150,6 +150,13 @@ with conn:
elif mode == "bad_rc":
response["result"]["ha"]["ocf_rc"] = 42
elif mode == "delegate":
response["result"] = {
"delegated": True,
"backend": "container",
"container_runtime": "podman",
}
conn.sendall(
json.dumps(response).encode("utf-8") + b"\n"
)
@@ -248,6 +255,7 @@ PY_REFUSED
echo "===== PROXY PROTOCOL CLASSIFICATION ====="
run_case valid valid 0
run_case delegation delegate 0
run_case missing_protocol missing_protocol 11
run_case wrong_protocol wrong_protocol 11
run_case wrong_resource wrong_resource 11