Error handling while running the Gluster commands

A new metric is added to capture the metrics collection errors
`glusterfs_error_count` with label "name" to identify the failed
metrics collector. Example:

```prometheus
glusterfs_error_count{name="list_volumes"} 1.0
glusterfs_error_count{name="list_peers"} 1.0
```

On error, volume and peer metrics will not be collected. Once
Glusterd comes back online or after the issue is fixed, then these
metrics will start appearing.

DW_FORM_data16 exception is fixed with the latest version
of the Crystal
(Ref: crystal-lang/crystal#12744)

Fixes: #38
Signed-off-by: Aravinda Vishwanathapura <[email protected]>

shard.lock

@@ -2,23 +2,23 @@ version: 2.0
 shards:
   backtracer:
     git: https://github.com/sija/backtracer.cr.git
-    version: 1.2.1
+    version: 1.2.2
 
   crometheus:
     git: https://github.com/darwinnn/crometheus.git
-    version: 0.3.0+git.commit.8394850abd90b976aa205f58de49a886c91aa802
+    version: 0.3.0+git.commit.c71a13174d02e3767b62faa1771132d4245e1ce5
 
   exception_page:
     git: https://github.com/crystal-loot/exception_page.git
-    version: 0.2.2
+    version: 0.4.1
 
   glustercli:
     git: https://github.com/aravindavk/glustercli-crystal.git
-    version: 0.2.0+git.commit.0881f270471be31d4f6d06f45d896e17adace0db
+    version: 0.2.0+git.commit.fb2f36881d79523990ca8b5f23212d4181718c8f
 
   kemal:
     git: https://github.com/kemalcr/kemal.git
-    version: 1.2.0
+    version: 1.5.0
 
   radix:
     git: https://github.com/luislavena/radix.git

src/metrics/errors.cr

@@ -0,0 +1,19 @@
+module GlusterMetricsExporter
+  Crometheus.alias ErrorGauge = Crometheus::Gauge[:name]
+
+  @@error_count = ErrorGauge.new(:error_count, "Metrics collection errors")
+
+  def self.clear_error_metrics
+    @@error_count.clear
+  end
+
+  handle_metrics(["error"]) do |metrics_data|
+    # Reset all Metrics to avoid stale data. Careful if
+    # counter type is used
+    clear_error_metrics
+
+    metrics_data.errors.each do |err|
+      @@error_count[name: err.name].set(1)
+    end
+  end
+end

src/metrics/helpers.cr

@@ -1,13 +1,23 @@
 require "glustercli"
 
 module GlusterMetricsExporter
+  struct MetricError
+    include JSON::Serializable
+
+    property name = ""
+
+    def initialize(@name)
+    end
+  end
+
   class MetricsData
     include JSON::Serializable
 
     property volumes = [] of GlusterCLI::VolumeInfo,
       peers = [] of GlusterCLI::NodeInfo,
       local_metrics = Hash(String, GlusterCLI::LocalMetrics).new,
-      exporter_health = Hash(String, Int32).new
+      exporter_health = Hash(String, Int32).new,
+      errors = [] of MetricError
 
     def self.collect
       data = MetricsData.new
@@ -24,11 +34,21 @@ module GlusterMetricsExporter
           status_collect = true
         end
 
-        data.volumes = cli.list_volumes(status: status_collect)
+        begin
+          data.volumes = cli.list_volumes(status: status_collect)
+        rescue ex : GlusterCLI::CommandException
+          data.errors << MetricError.new("list_volumes")
+          Log.error &.emit("Error while collecting the Volumes metrics", error: ex.message)
+        end
       end
 
       if GlusterMetricsExporter.config.enabled?("peer")
-        data.peers = cli.list_peers
+        begin
+          data.peers = cli.list_peers
+        rescue ex : GlusterCLI::CommandException
+          data.errors << MetricError.new("list_peers")
+          Log.error &.emit("Error while collecting the Peers metrics", error: ex.message)
+        end
       end
 
       # TODO: API calls concurrently

src/metrics/peer.cr

@@ -20,7 +20,7 @@ module GlusterMetricsExporter
 
     metrics_data.peers.each do |peer|
       # Peer State 1 => Connected, 0 => Disconnected/Unknown
-      state = peer.connected ? 1 : 0
+      state = peer.connected? ? 1 : 0
       @@peer_state[hostname: peer.hostname].set(state)
     end
   end