Use local vars, not instance vars, in metricsHandler

[dmazin]

When Falcon starts up, it creates an individual instance of metricsHandler for each metric type:

def falcon_app():
    api = falcon.API()
    api.add_route("/health", metricsHandler(config, metrics_type="health"))

The same metricsHandler instance is used for every /health request.

This, by itself, is not an issue. However, because the instance stores state (such as IP/hostname) in instances variables, these can get clobbered if we process multiple requests simultaneously.

By instance variables I mean usage of self:

class metricsHandler:
    ...
    def on_get(self, req, resp):
        self.target = req.get_param("target")

Here is you can reproduce this.
First, let's add some logging to show the issue.

diff --git a/collector.py b/collector.py
index 32d29d9..e183ddc 100644
--- a/collector.py
+++ b/collector.py
@@ -20,6 +20,8 @@ class RedfishMetricsCollector(object):
        self.target = target
        self.host = host

+       logging.info(f"**** Initializing RedfishMetricsCollector for hostname {self.host} and IP {self.target}")
+
        self._username = usr
        self._password = pwd

diff --git a/handler.py b/handler.py
index b035449..1627ffc 100644
--- a/handler.py
+++ b/handler.py
@@ -33,6 +33,8 @@ class metricsHandler:
        self.metrics_type = metrics_type

    def on_get(self, req, resp):
+       logging.info(f"**** The id of this metricsHandler instance is {id(self)}")
+
        self.target = req.get_param("target")
        if not self.target:
            logging.error("No target parameter provided!")
@@ -87,6 +89,7 @@ class metricsHandler:
        logging.debug(f"{self.target}: Using user {usr}")

+      logging.info(f"**** Scraping {self.host}, the client asked for {req.get_param('target')}")
        with RedfishMetricsCollector(
            self._config,
            target = self.target,

I am going to use the following bash script to quickly make two requests against the exporter:

#!/bin/bash
idracs=("foo" "bar")
for idrac in "${idracs[@]}"; do
    curl "http://localhost:9229/health?target=${idrac}&job=redfish" -o "${idrac}_stats.txt" &
done

Now, let’s observe the logs:

handler.py:36 INFO  *** The id of this metricsHandler instance is 131570941555444
handler.py:36 INFO  *** The id of this metricsHandler instance is 131570941555444
handler.py:94 INFO  *** Scraping foo; the client asked for foo
handler.py:94 INFO  *** Scraping foo; the client asked for bar
collector.py:23 INFO  *** Initializing RedfishMetricsCollector for hostname foo and IP 10.0.0.3.
collector.py:23 INFO  *** Initializing RedfishMetricsCollector for hostname foo and IP 10.0.0.3.

We see a few things:
• Lines 1,2: Indeed, the very same metricsHandler instance processes both requests (this is evidenced by the id being identical). Again, this isn’t an issue unto itself; just saying it’s the same instance so we can’t use self safely.
• Lines 3,4: You can see that it scrapes foo even though the client asked for bar
• Lines 5,6: From collector.py, you can see that the collector is initialized for foo and IP 10.0.0.3 twice even though we had asked for bar on the second request.

fixes #9