feat(ops): add systemd checks for ollama and llama.cpp
This commit is contained in:
@@ -13,6 +13,10 @@ MCP_URL="${MCP_URL:-http://192.168.153.113:18802/mcp}"
|
||||
OLLAMA_URL="${OLLAMA_URL:-http://127.0.0.1:11434}"
|
||||
OLLAMA_EMBED_MODEL="${OLLAMA_EMBED_MODEL:-nomic-embed-text:latest}"
|
||||
OLLAMA_REQUIRED="${OLLAMA_REQUIRED:-false}"
|
||||
OLLAMA_SERVICE="${OLLAMA_SERVICE:-ollama.service}"
|
||||
LLAMA_CPP_SERVICE="${LLAMA_CPP_SERVICE:-llama-server.service}"
|
||||
LLAMA_CPP_URL="${LLAMA_CPP_URL:-http://127.0.0.1:8081/health}"
|
||||
LLAMA_CPP_REQUIRED="${LLAMA_CPP_REQUIRED:-false}"
|
||||
MINIO_URL="${MINIO_URL:-http://192.168.153.253:9000}"
|
||||
DOCKER_EXPECTED_CONTAINERS="${DOCKER_EXPECTED_CONTAINERS:-searxng whisper-server brave-search}"
|
||||
DOCKER_REQUIRE_EXPECTED="${DOCKER_REQUIRE_EXPECTED:-false}"
|
||||
@@ -128,7 +132,28 @@ else
|
||||
add_next "Create backup log or fix backup script path"
|
||||
fi
|
||||
|
||||
# 3) Docker service health (containers + health state)
|
||||
# 3) systemd service health (local model runtimes)
|
||||
if command -v systemctl >/dev/null 2>&1; then
|
||||
for unit in "$OLLAMA_SERVICE" "$LLAMA_CPP_SERVICE"; do
|
||||
if systemctl is-active --quiet "$unit"; then
|
||||
add_watch "P4 systemd service active: ${unit}"
|
||||
else
|
||||
if [[ "$unit" == "$OLLAMA_SERVICE" && "${OLLAMA_REQUIRED}" != "true" ]]; then
|
||||
add_watch "P3 systemd service inactive: ${unit} (optional)"
|
||||
elif [[ "$unit" == "$LLAMA_CPP_SERVICE" && "${LLAMA_CPP_REQUIRED}" != "true" ]]; then
|
||||
add_watch "P3 systemd service inactive: ${unit} (optional)"
|
||||
else
|
||||
add_now "P1 systemd service inactive: ${unit}"
|
||||
mark_p1
|
||||
add_next "Check unit: systemctl status ${unit}"
|
||||
fi
|
||||
fi
|
||||
done
|
||||
else
|
||||
add_watch "P3 systemctl not available"
|
||||
fi
|
||||
|
||||
# 4) Docker service health (containers + health state)
|
||||
if command -v docker >/dev/null 2>&1; then
|
||||
if docker ps --format '{{.Names}}\t{{.Status}}' >"$ARTIFACT_DIR/docker-ps.txt" 2>"$ARTIFACT_DIR/docker-ps.err"; then
|
||||
for svc in $DOCKER_EXPECTED_CONTAINERS; do
|
||||
@@ -159,13 +184,25 @@ else
|
||||
add_watch "P3 docker CLI not available"
|
||||
fi
|
||||
|
||||
# 4) Key LAN + local service probes
|
||||
# 5) Key LAN + local service probes
|
||||
http_probe "searxng" "$SEARX_URL" '^200$'
|
||||
http_probe "whisper" "$WHISPER_URL" '^200$'
|
||||
http_probe "brave-mcp" "$MCP_URL" '^(200|406)$'
|
||||
http_probe "minio-live" "${MINIO_URL%/}/minio/health/live" '^200$'
|
||||
|
||||
# 5) Ollama embeddings availability + target model
|
||||
if curl -sS -m 6 "$LLAMA_CPP_URL" >"$ARTIFACT_DIR/llamacpp-health.txt" 2>"$ARTIFACT_DIR/llamacpp-health.err"; then
|
||||
add_watch "P4 llama.cpp endpoint responsive (${LLAMA_CPP_URL})"
|
||||
else
|
||||
if [[ "${LLAMA_CPP_REQUIRED}" == "true" ]]; then
|
||||
add_now "P1 llama.cpp endpoint unreachable (${LLAMA_CPP_URL})"
|
||||
mark_p1
|
||||
add_next "Check llama.cpp bind address/port and service logs"
|
||||
else
|
||||
add_watch "P3 llama.cpp endpoint unreachable at configured URL (${LLAMA_CPP_URL})"
|
||||
fi
|
||||
fi
|
||||
|
||||
# 6) Ollama embeddings availability + target model
|
||||
if curl -sS -m 6 "${OLLAMA_URL%/}/api/tags" >"$ARTIFACT_DIR/ollama-tags.json" 2>"$ARTIFACT_DIR/ollama-tags.err"; then
|
||||
if jq -e --arg model "$OLLAMA_EMBED_MODEL" '.models[]? | select(.name == $model)' "$ARTIFACT_DIR/ollama-tags.json" >/dev/null 2>&1; then
|
||||
add_watch "P4 Ollama up; embedding model present (${OLLAMA_EMBED_MODEL})"
|
||||
@@ -184,7 +221,7 @@ else
|
||||
fi
|
||||
fi
|
||||
|
||||
# 6) Host pressure: disk + memory
|
||||
# 7) Host pressure: disk + memory
|
||||
root_disk_pct="$(df -P / | awk 'NR==2 {gsub(/%/,"",$5); print $5}' 2>/dev/null || echo 0)"
|
||||
if [[ "$root_disk_pct" =~ ^[0-9]+$ ]]; then
|
||||
if (( root_disk_pct >= 95 )); then
|
||||
|
||||
Reference in New Issue
Block a user