Tier2メトリクス(コマンド実行数・レイテンシ・エラー率)を追加

bot.py

@@ -11,13 +11,23 @@
 import wikidata.client
 import asyncio
 import datetime
+import time
 import traceback
 
 import aiohttp
 import aiosqlite
 import database
 
-from metrics import start_metrics_server, BOT_READY, BOT_GUILD_COUNT, BOT_LATENCY_SECONDS, BOT_SHARD_COUNT
+from metrics import (
+    start_metrics_server,
+    BOT_READY,
+    BOT_GUILD_COUNT,
+    BOT_LATENCY_SECONDS,
+    BOT_SHARD_COUNT,
+    BOT_COMMANDS_TOTAL,
+    BOT_COMMAND_ERRORS_TOTAL,
+    BOT_COMMAND_LATENCY_SECONDS,
+)
 
 from twitter import *
 from dateutil.relativedelta import relativedelta as rdelta
@@ -649,6 +659,7 @@ async def rtopic(ctx, ch:discord.TextChannel=None):
 
 @bot.event
 async def on_command(ctx:commands.Context):
+    ctx._t2_start = time.monotonic()
     if ctx.interaction:return
     ch = await bot.fetch_channel(693048961107230811)
     e = discord.Embed(title=f"prefixコマンド:{ctx.command.full_parent_name}の実行",
@@ -667,6 +678,15 @@ async def on_command(ctx:commands.Context):
     e.timestamp = ctx.message.created_at
     await ch.send(embed=e)
 
+@bot.event
+async def on_command_completion(ctx:commands.Context):
+    command_name = ctx.command.qualified_name
+    cmd_type = "slash" if ctx.interaction else "prefix"
+    BOT_COMMANDS_TOTAL.labels(command=command_name, type=cmd_type).inc()
+    start = getattr(ctx, "_t2_start", None)
+    if start is not None:
+        BOT_COMMAND_LATENCY_SECONDS.labels(command=command_name, type=cmd_type).observe(time.monotonic() - start)
+
 @bot.event
 async def on_interaction(interaction:discord.Interaction):
     ch = await bot.fetch_channel(693048961107230811)
@@ -698,6 +718,9 @@ async def on_command_error(ctx, error):
     el"""
     if isinstance(error, commands.HybridCommandError):
         error = error.original
+    command_name = ctx.command.qualified_name if ctx.command else "(unknown)"
+    cmd_type = "slash" if ctx.interaction else "prefix"
+    BOT_COMMAND_ERRORS_TOTAL.labels(command=command_name, type=cmd_type, error_type=type(error).__name__).inc()
     if isinstance(error, (commands.CommandOnCooldown, app_commands.CommandOnCooldown)):
         # クールダウン
         embed = discord.Embed(title=await ctx._("cmd-error-t"), description=await ctx._(

metrics.py

@@ -1,10 +1,26 @@
-from prometheus_client import start_http_server, Gauge
+from prometheus_client import start_http_server, Gauge, Counter, Histogram
 
 BOT_READY = Gauge("sina_chan_ready", "1 if the bot is ready, 0 otherwise")
 BOT_GUILD_COUNT = Gauge("sina_chan_guild_count", "Number of guilds the bot is in")
 BOT_LATENCY_SECONDS = Gauge("sina_chan_latency_seconds", "Discord gateway latency")
 BOT_SHARD_COUNT = Gauge("sina_chan_shard_count", "Number of shards")
 
+BOT_COMMANDS_TOTAL = Counter(
+    "sina_chan_commands_total",
+    "Number of commands that completed successfully",
+    ["command", "type"],
+)
+BOT_COMMAND_ERRORS_TOTAL = Counter(
+    "sina_chan_command_errors_total",
+    "Number of commands that ended in an error",
+    ["command", "type", "error_type"],
+)
+BOT_COMMAND_LATENCY_SECONDS = Histogram(
+    "sina_chan_command_latency_seconds",
+    "Command execution time in seconds, from invocation to successful completion",
+    ["command", "type"],
+)
+
 
 def start_metrics_server(port: int = 9200) -> None:
     # ループバック限定。既存のmysqld_exporter・blackbox_exporterと同じ考え方で、