managing-grpc

Use when working with Grpc — gRPC service management - service discovery via reflection, health checking, load balancing analysis, channelz diagnostics, and proto schema inspection. Use when debugging

SKILL.md

GitHub

gRPC Management Skill

Manage and analyze gRPC services using reflection, health checks, channelz, and grpcurl.

Core Helper Functions

#!/bin/bash

# gRPC target
GRPC_TARGET="${GRPC_TARGET:-localhost:50051}"
GRPC_PLAINTEXT="${GRPC_PLAINTEXT:---plaintext}"

# grpcurl wrapper
grpc_call() {
    grpcurl ${GRPC_PLAINTEXT} "$@" "$GRPC_TARGET" 2>&1
}

# List services via reflection
grpc_services() {
    grpc_call -list 2>/dev/null || echo "ERROR: Reflection not enabled or server unreachable"
}

# Describe a service or method
grpc_describe() {
    local symbol="$1"
    grpc_call -describe "$symbol" 2>/dev/null
}

MANDATORY: Discovery-First Pattern

Always discover available services and check health before performing operations.

Phase 1: Discovery

#!/bin/bash

echo "=== Server Connectivity ==="
grpc_call -connect-timeout 5 grpc.health.v1.Health/Check 2>&1 | head -5
echo ""

echo "=== Available Services ==="
grpc_services

echo ""
echo "=== Service Details ==="
for svc in $(grpc_services | grep -v "^grpc\.\|^ERROR"); do
    echo "--- ${svc} ---"
    grpc_describe "$svc" | head -20
done

echo ""
echo "=== Health Status ==="
grpc_call grpc.health.v1.Health/Check 2>/dev/null \
    || echo "Health check service not available"

Output Rules

  • TOKEN EFFICIENCY: Target <=50 lines per output
  • Summarize proto definitions; do not dump full message schemas
  • Use grpcurl's JSON output format for structured data

Common Operations

Service Discovery and Inspection

#!/bin/bash

echo "=== All Services and Methods ==="
for svc in $(grpc_services | grep -v "^grpc\.\|^ERROR"); do
    echo "Service: ${svc}"
    grpc_describe "$svc" | grep "rpc " | sed 's/^/  /'
done

echo ""
echo "=== Message Types ==="
SERVICE="${1:-}"
if [ -n "$SERVICE" ]; then
    for method in $(grpc_describe "$SERVICE" | grep "rpc " | awk '{print $2}' | tr -d '('); do
        echo "--- ${method} ---"
        # Extract input/output types
        grpc_describe "${SERVICE}.${method}" 2>/dev/null | head -5
    done
fi

echo ""
echo "=== Proto File Reconstruction ==="
grpc_call -describe-all 2>/dev/null | head -50

Health Checking

#!/bin/bash

echo "=== Overall Health ==="
grpc_call grpc.health.v1.Health/Check 2>/dev/null

echo ""
echo "=== Per-Service Health ==="
for svc in $(grpc_services | grep -v "^grpc\.\|^ERROR"); do
    status=$(grpc_call -d "{\"service\": \"${svc}\"}" grpc.health.v1.Health/Check 2>/dev/null | jq -r '.status // "UNKNOWN"')
    echo "${svc}: ${status}"
done

echo ""
echo "=== Health Watch (snapshot) ==="
timeout 3 grpcurl ${GRPC_PLAINTEXT} -d '{"service": ""}' "$GRPC_TARGET" grpc.health.v1.Health/Watch 2>/dev/null | head -10

Load Balancing Analysis

#!/bin/bash

echo "=== DNS Resolution for Target ==="
HOST=$(echo "$GRPC_TARGET" | cut -d: -f1)
PORT=$(echo "$GRPC_TARGET" | cut -d: -f2)
echo "Target: ${HOST}:${PORT}"
dig +short "$HOST" 2>/dev/null || nslookup "$HOST" 2>/dev/null | grep "Address" | tail -n+2

echo ""
echo "=== Connection Test to Multiple Backends ==="
for ip in $(dig +short "$HOST" 2>/dev/null); do
    echo "--- ${ip}:${PORT} ---"
    grpcurl ${GRPC_PLAINTEXT} -connect-timeout 3 "${ip}:${PORT}" grpc.health.v1.Health/Check 2>/dev/null \
        && echo "  Status: REACHABLE" || echo "  Status: UNREACHABLE"
done

echo ""
echo "=== Load Balancing Config (if using xDS) ==="
echo "Check service mesh or load balancer configuration for gRPC-aware routing."
echo "Common LB policies: round_robin, pick_first, grpclb, xds"

Channelz Diagnostics

#!/bin/bash

echo "=== Top Channels ==="
grpc_call grpc.channelz.v1.Channelz/GetTopChannels 2>/dev/null \
    | jq '{channels: [.channel[] | {
        ref: .ref.name,
        state: .data.state.state,
        target: .data.target,
        calls_started: .data.callsStarted,
        calls_succeeded: .data.callsSucceeded,
        calls_failed: .data.callsFailed,
        last_call: .data.lastCallStartedTimestamp
    }]}' 2>/dev/null || echo "Channelz not enabled on this server"

echo ""
echo "=== Servers ==="
grpc_call grpc.channelz.v1.Channelz/GetServers 2>/dev/null \
    | jq '{servers: [.server[] | {
        id: .ref.serverId,
        calls_started: .data.callsStarted,
        calls_succeeded: .data.callsSucceeded,
        calls_failed: .data.callsFailed,
        last_call: .data.lastCallStartedTimestamp
    }]}' 2>/dev/null || echo "Channelz not available"

echo ""
echo "=== Socket Details ==="
grpc_call grpc.channelz.v1.Channelz/GetTopChannels 2>/dev/null \
    | jq '[.channel[].subchannel[]?.ref.subchannelId // empty]' 2>/dev/null | head -10

Method Invocation (Read-Only)

#!/bin/bash
SERVICE="${1:?Service name required}"
METHOD="${2:?Method name required}"

echo "=== Method Description ==="
grpc_describe "${SERVICE}/${METHOD}"

echo ""
echo "=== Input Message Schema ==="
input_type=$(grpc_describe "${SERVICE}/${METHOD}" | grep "input type" | awk '{print $NF}' | tr -d '.')
if [ -n "$input_type" ]; then
    grpc_describe "$input_type" 2>/dev/null
fi

echo ""
echo "=== Dry Run (empty request) ==="
echo "Command: grpcurl ${GRPC_PLAINTEXT} -d '{}' ${GRPC_TARGET} ${SERVICE}/${METHOD}"
echo "NOTE: Only execute after confirming the method is read-only (not a mutating RPC)"

Safety Rules

  • Read-only by default: Only use reflection, health checks, and channelz for inspection
  • Never invoke mutating RPCs (Create, Update, Delete methods) without explicit user confirmation
  • Identify method semantics: Check method names and descriptions before invocation; treat unknown methods as potentially mutating
  • Connection limits: Avoid opening excessive connections; reuse grpcurl sessions where possible
  • TLS verification: In production, never use --plaintext; always verify certificates

Output Format

Present results as a structured report:

Managing Grpc Report
════════════════════
Resources discovered: [count]

Resource       Status    Key Metric    Issues
──────────────────────────────────────────────
[name]         [ok/warn] [value]       [findings]

Summary: [total] resources | [ok] healthy | [warn] warnings | [crit] critical
Action Items: [list of prioritized findings]

Target ≤50 lines of output. Use tables for multi-resource comparisons.

Anti-Hallucination Rules

  1. NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
  2. NEVER fabricate metric names or dimensions — verify against the service documentation or --help output.
  3. NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
  4. ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
  5. ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.

Counter-Rationalizations

Shortcut Counter Why
"I'll skip discovery and check known resources" Always run Phase 1 discovery first Resource names change, new resources appear — assumed names cause errors
"The user only asked for a quick check" Follow the full discovery → analysis flow Quick checks miss critical issues; structured analysis catches silent failures
"Default configuration is probably fine" Audit configuration explicitly Defaults often leave logging, security, and optimization features disabled
"Metrics aren't needed for this" Always check relevant metrics when available API/CLI responses show current state; metrics reveal trends and intermittent issues
"I don't have access to that" Try the command and report the actual error Assumed permission failures prevent useful investigation; actual errors are informative

Common Pitfalls

  • Reflection disabled: Many production servers disable reflection for security; use proto files instead
  • Streaming RPCs: Server-streaming and bidirectional RPCs require special handling with grpcurl (-d @ for client streaming)
  • Deadline/timeout: Always set deadlines; gRPC calls without deadlines can hang indefinitely
  • Status codes: gRPC status codes differ from HTTP; UNAVAILABLE (14) means transient failure, NOT_FOUND (5) means the resource does not exist
  • Binary encoding: gRPC uses protobuf binary format; raw packet inspection requires protobuf decoding tools