From 74c5c00305f6a047900a5ed2ff039cbfb066a398 Mon Sep 17 00:00:00 2001
From: Benjamin Schanzel <benjamin.schanzel@bmw.de>
Date: Tue, 4 Jan 2022 14:12:30 +0100
Subject: [PATCH] Export current tenant limit stats

This adds a new statsd gauge which, in addition to the existing provider
limits, exports the currently configured tenant limits. This is in the
form ``nodepool.tenant_limits.TENANT.[cores,ram,instances]``.

Change-Id: I8e10a0974210d25d071dbbd63849a921fc8b79a2
---
 doc/source/operation.rst | 11 +++++++++++
 nodepool/launcher.py     |  2 ++
 nodepool/stats.py        | 24 ++++++++++++++++++++++--
 3 files changed, 35 insertions(+), 2 deletions(-)

diff --git a/doc/source/operation.rst b/doc/source/operation.rst
index fa79e419f..db21fc02a 100644
--- a/doc/source/operation.rst
+++ b/doc/source/operation.rst
@@ -495,6 +495,17 @@ The following metrics are produced by a ``nodepool-launcher`` process:
    Number of nodes with a specific label in a specific state. See
    :ref:`nodepool.nodes <nodepool_nodes>` for a list of possible states.
 
+.. zuul:stat:: nodepool.tenant_limits.<tenant>.<limit>
+   :type: guage
+
+   The currently configured resource limits of a tenant.
+
+   limit can be:
+
+   * cores
+   * instances
+   * ram
+
 Provider Metrics
 ^^^^^^^^^^^^^^^^
 
diff --git a/nodepool/launcher.py b/nodepool/launcher.py
index c6f619f6e..7e80da253 100644
--- a/nodepool/launcher.py
+++ b/nodepool/launcher.py
@@ -338,6 +338,8 @@ class PoolWorker(threading.Thread, stats.StatsReporter):
 
                 self.updateProviderLimits(
                     self.nodepool.config.providers.get(self.provider_name))
+                self.updateTenantLimits(
+                    self.nodepool.config.tenant_resource_limits)
 
                 if not self.paused_handler:
                     while not self._assignHandlers():
diff --git a/nodepool/stats.py b/nodepool/stats.py
index 6331e097a..229a18d8b 100644
--- a/nodepool/stats.py
+++ b/nodepool/stats.py
@@ -41,6 +41,10 @@ def get_client():
         return None
 
 
+def normalize_statsd_name(name):
+    return name.replace('.', '_').replace(':', '_')
+
+
 class StatsReporter(object):
     '''
     Class adding statsd reporting functionality.
@@ -72,8 +76,7 @@ class StatsReporter(object):
         if self.handler.request.requestor:
             # Replace '.' which is a graphite hierarchy, and ':' which is
             # a statsd delimeter.
-            requestor = self.handler.request.requestor.replace('.', '_')
-            requestor = requestor.replace(':', '_')
+            requestor = normalize_statsd_name(self.handler.request.requestor)
             keys.append('nodepool.launch.requestor.%s.%s' %
                         (requestor, subkey))
 
@@ -159,3 +162,20 @@ class StatsReporter(object):
         pipeline.gauge(key, max_servers)
 
         pipeline.send()
+
+    def updateTenantLimits(self, tenant_limits):
+        if not self._statsd:
+            return
+
+        pipeline = self._statsd.pipeline()
+        # nodepool.tenant_limits.TENANT.[cores,ram,instances]
+        key_template = 'nodepool.tenant_limits.%s.%s'
+        for tenant, limits in tenant_limits.items():
+            for k, lim in limits.items():
+                # normalize for statsd name, as parts come from arbitrary
+                # user config
+                tenant = normalize_statsd_name(tenant)
+                k = normalize_statsd_name(k)
+                key = key_template % (tenant, k)
+                pipeline.gauge(key, lim)
+        pipeline.send()