Metrics for IO errors on Mac. (#1636)

* Metrics for IO errors and retries on Mac.

Signed-off-by: Tom Wilkie <tom@grafana.com>

CHANGELOG.md

@@ -3,6 +3,7 @@
 * [CHANGE]
 * [FEATURE]
 * [ENHANCEMENT] Add model_name and stepping to node_cpu_info metric #1617
+* [ENHANCEMENT] Add metrics for IO errors and retires on Darwin. #1636
 * [BUGFIX]
 
 ## 1.0.0-rc.0 / 2020-02-20

collector/diskstats_darwin.go

@@ -125,6 +125,62 @@ func NewDiskstatsCollector(logger log.Logger) (Collector, error) {
 					return float64(stat.BytesWritten)
 				},
 			},
+			{
+				typedDesc: typedDesc{
+					desc: prometheus.NewDesc(
+						prometheus.BuildFQName(namespace, diskSubsystem, "read_errors_total"),
+						"The total number of read errors.",
+						diskLabelNames,
+						nil,
+					),
+					valueType: prometheus.CounterValue,
+				},
+				value: func(stat *iostat.DriveStats) float64 {
+					return float64(stat.ReadErrors)
+				},
+			},
+			{
+				typedDesc: typedDesc{
+					desc: prometheus.NewDesc(
+						prometheus.BuildFQName(namespace, diskSubsystem, "write_errors_total"),
+						"The total number of write errors.",
+						diskLabelNames,
+						nil,
+					),
+					valueType: prometheus.CounterValue,
+				},
+				value: func(stat *iostat.DriveStats) float64 {
+					return float64(stat.WriteErrors)
+				},
+			},
+			{
+				typedDesc: typedDesc{
+					desc: prometheus.NewDesc(
+						prometheus.BuildFQName(namespace, diskSubsystem, "read_retries_total"),
+						"The total number of read retries.",
+						diskLabelNames,
+						nil,
+					),
+					valueType: prometheus.CounterValue,
+				},
+				value: func(stat *iostat.DriveStats) float64 {
+					return float64(stat.ReadRetries)
+				},
+			},
+			{
+				typedDesc: typedDesc{
+					desc: prometheus.NewDesc(
+						prometheus.BuildFQName(namespace, diskSubsystem, "write_retries_total"),
+						"The total number of write retries.",
+						diskLabelNames,
+						nil,
+					),
+					valueType: prometheus.CounterValue,
+				},
+				value: func(stat *iostat.DriveStats) float64 {
+					return float64(stat.WriteRetries)
+				},
+			},
 		},
 		logger: logger,
 	}, nil

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/godbus/dbus v0.0.0-20190402143921-271e53dc4968
 	github.com/golang/protobuf v1.3.3 // indirect
 	github.com/hodgesds/perf-utils v0.0.8
-	github.com/lufia/iostat v0.0.0-20170605150913-9f7362b77ad3
+	github.com/lufia/iostat v1.1.0
 	github.com/mattn/go-xmlrpc v0.0.3
 	github.com/mdlayher/genetlink v1.0.0 // indirect
 	github.com/mdlayher/netlink v1.1.0 // indirect

go.sum

@@ -174,8 +174,8 @@ github.com/kr/text v0.1.0 h1:45sCR5RtlFHMR4UwH9sdQ5TC8v0qDQCHnXt+kaKSTVE=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/lightstep/lightstep-tracer-common/golang/gogo v0.0.0-20190605223551-bc2310a04743/go.mod h1:qklhhLq1aX+mtWk9cPHPzaBjWImj5ULL6C7HFJtXQMM=
 github.com/lightstep/lightstep-tracer-go v0.18.1/go.mod h1:jlF1pusYV4pidLvZ+XD0UBX0ZE6WURAspgAczcDHrL4=
-github.com/lufia/iostat v0.0.0-20170605150913-9f7362b77ad3 h1:XGhvld9vIpj929Gri5ybjukYZeyZwKkFkqgATqBQiOs=
-github.com/lufia/iostat v0.0.0-20170605150913-9f7362b77ad3/go.mod h1:lRgtFVamD7L7GaXOSwBiuXMwU3Aicfn5h66LVs4u2SA=
+github.com/lufia/iostat v1.1.0 h1:Z1wa4Hhxwi8uSKfgRsFc5RLtt3SuFPIOgkiPGkUtHDY=
+github.com/lufia/iostat v1.1.0/go.mod h1:rEPNA0xXgjHQjuI5Cy05sLlS2oRcSlWHRLrvh/AQ+Pg=
 github.com/lyft/protoc-gen-validate v0.0.13/go.mod h1:XbGvPuh87YZc5TdIa2/I4pLk0QoUACkjt2znoq26NVQ=
 github.com/mattn/go-colorable v0.0.9/go.mod h1:9vuHe8Xs5qXnSaW/c/ABM9alt+Vo+STaOChaDxuIBZU=
 github.com/mattn/go-isatty v0.0.3/go.mod h1:M+lRXTBqGeGNdLjl/ufCoiOlB5xdOkqRJdNxMWT7Zi4=

vendor/github.com/lufia/iostat/go.mod

@@ -0,0 +1,3 @@
+module github.com/lufia/iostat
+
+go 1.14

vendor/github.com/lufia/iostat/iostat.go

@@ -17,6 +17,10 @@ type DriveStats struct {
 	TotalWriteTime time.Duration
 	ReadLatency    time.Duration
 	WriteLatency   time.Duration
+	ReadErrors     int64
+	WriteErrors    int64
+	ReadRetries    int64
+	WriteRetries   int64
 }
 
 // CPUStats represents CPU statistics.

vendor/github.com/lufia/iostat/iostat_darwin.c

@@ -95,6 +95,10 @@ static struct {
 	{kIOBlockStorageDriverStatisticsTotalWriteTimeKey, offsetof(DriveStats, writetime)},
 	{kIOBlockStorageDriverStatisticsLatentReadTimeKey, offsetof(DriveStats, readlat)},
 	{kIOBlockStorageDriverStatisticsLatentWriteTimeKey, offsetof(DriveStats, writelat)},
+	{kIOBlockStorageDriverStatisticsReadErrorsKey, offsetof(DriveStats, readerrs)},
+	{kIOBlockStorageDriverStatisticsWriteErrorsKey, offsetof(DriveStats, writeerrs)},
+	{kIOBlockStorageDriverStatisticsReadRetriesKey, offsetof(DriveStats, readretries)},
+	{kIOBlockStorageDriverStatisticsWriteRetriesKey, offsetof(DriveStats, writeretries)},
 };
 
 static int

vendor/github.com/lufia/iostat/iostat_darwin.go

@@ -32,6 +32,10 @@ func ReadDriveStats() ([]*DriveStats, error) {
 			TotalWriteTime: time.Duration(buf[i].writetime),
 			ReadLatency:    time.Duration(buf[i].readlat),
 			WriteLatency:   time.Duration(buf[i].writelat),
+			ReadErrors:     int64(buf[i].readerrs),
+			WriteErrors:    int64(buf[i].writeerrs),
+			ReadRetries:    int64(buf[i].readretries),
+			WriteRetries:   int64(buf[i].writeretries),
 		}
 	}
 	return stats, nil

vendor/github.com/lufia/iostat/iostat_darwin.h

@@ -19,6 +19,10 @@ struct DriveStats {
 	int64_t writetime;
 	int64_t readlat;
 	int64_t writelat;
+	int64_t readerrs;
+	int64_t writeerrs;
+	int64_t readretries;
+	int64_t writeretries;
 };
 
 struct CPUStats {

vendor/modules.txt

@@ -30,7 +30,7 @@ github.com/golang/protobuf/ptypes/duration
 github.com/golang/protobuf/ptypes/timestamp
 # github.com/hodgesds/perf-utils v0.0.8
 github.com/hodgesds/perf-utils
-# github.com/lufia/iostat v0.0.0-20170605150913-9f7362b77ad3
+# github.com/lufia/iostat v1.1.0
 github.com/lufia/iostat
 # github.com/mattn/go-xmlrpc v0.0.3
 github.com/mattn/go-xmlrpc