1 // Copyright 2020 The Go Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style
3 // license that can be found in the LICENSE file.
5 // Note: run 'go generate' (which will run 'go test -generate') to update the "Supported metrics" list.
6 //go:generate go test -run=Docs -generate
9 Package metrics provides a stable interface to access implementation-defined
10 metrics exported by the Go runtime. This package is similar to existing functions
11 like [runtime.ReadMemStats] and [debug.ReadGCStats], but significantly more general.
13 The set of metrics defined by this package may evolve as the runtime itself
14 evolves, and also enables variation across Go implementations, whose relevant
15 metric sets may not intersect.
19 Metrics are designated by a string key, rather than, for example, a field name in
20 a struct. The full list of supported metrics is always available in the slice of
21 Descriptions returned by All. Each Description also includes useful information
24 Thus, users of this API are encouraged to sample supported metrics defined by the
25 slice returned by All to remain compatible across Go versions. Of course, situations
26 arise where reading specific metrics is critical. For these cases, users are
27 encouraged to use build tags, and although metrics may be deprecated and removed,
28 users should consider this to be an exceptional and rare event, coinciding with a
29 very large change in a particular Go implementation.
31 Each metric key also has a "kind" that describes the format of the metric's value.
32 In the interest of not breaking users of this package, the "kind" for a given metric
33 is guaranteed not to change. If it must change, then a new metric will be introduced
34 with a new key and a new "kind."
38 As mentioned earlier, metric keys are strings. Their format is simple and well-defined,
39 designed to be both human and machine readable. It is split into two components,
40 separated by a colon: a rooted path and a unit. The choice to include the unit in
41 the key is motivated by compatibility: if a metric's unit changes, its semantics likely
42 did also, and a new key should be introduced.
44 For more details on the precise definition of the metric key's path and unit formats, see
45 the documentation of the Name field of the Description struct.
49 This package supports metrics whose values have a floating-point representation. In
50 order to improve ease-of-use, this package promises to never produce the following
51 classes of floating-point values: NaN, infinity.
55 Below is the full list of supported metrics, ordered lexicographically.
57 /cgo/go-to-c-calls:calls
58 Count of calls made from Go to C by the current process.
60 /cpu/classes/gc/mark/assist:cpu-seconds
61 Estimated total CPU time goroutines spent performing GC
62 tasks to assist the GC and prevent it from falling behind the
63 application. This metric is an overestimate, and not directly
64 comparable to system CPU time measurements. Compare only with
65 other /cpu/classes metrics.
67 /cpu/classes/gc/mark/dedicated:cpu-seconds
68 Estimated total CPU time spent performing GC tasks on processors
69 (as defined by GOMAXPROCS) dedicated to those tasks. This
70 includes time spent with the world stopped due to the GC. This
71 metric is an overestimate, and not directly comparable to system
72 CPU time measurements. Compare only with other /cpu/classes
75 /cpu/classes/gc/mark/idle:cpu-seconds
76 Estimated total CPU time spent performing GC tasks on spare CPU
77 resources that the Go scheduler could not otherwise find a use
78 for. This should be subtracted from the total GC CPU time to
79 obtain a measure of compulsory GC CPU time. This metric is an
80 overestimate, and not directly comparable to system CPU time
81 measurements. Compare only with other /cpu/classes metrics.
83 /cpu/classes/gc/pause:cpu-seconds
84 Estimated total CPU time spent with the application paused by
85 the GC. Even if only one thread is running during the pause,
86 this is computed as GOMAXPROCS times the pause latency because
87 nothing else can be executing. This is the exact sum of samples
88 in /gc/pause:seconds if each sample is multiplied by GOMAXPROCS
89 at the time it is taken. This metric is an overestimate,
90 and not directly comparable to system CPU time measurements.
91 Compare only with other /cpu/classes metrics.
93 /cpu/classes/gc/total:cpu-seconds
94 Estimated total CPU time spent performing GC tasks. This metric
95 is an overestimate, and not directly comparable to system CPU
96 time measurements. Compare only with other /cpu/classes metrics.
97 Sum of all metrics in /cpu/classes/gc.
99 /cpu/classes/idle:cpu-seconds
100 Estimated total available CPU time not spent executing
101 any Go or Go runtime code. In other words, the part of
102 /cpu/classes/total:cpu-seconds that was unused. This metric is
103 an overestimate, and not directly comparable to system CPU time
104 measurements. Compare only with other /cpu/classes metrics.
106 /cpu/classes/scavenge/assist:cpu-seconds
107 Estimated total CPU time spent returning unused memory to the
108 underlying platform in response eagerly in response to memory
109 pressure. This metric is an overestimate, and not directly
110 comparable to system CPU time measurements. Compare only with
111 other /cpu/classes metrics.
113 /cpu/classes/scavenge/background:cpu-seconds
114 Estimated total CPU time spent performing background tasks to
115 return unused memory to the underlying platform. This metric is
116 an overestimate, and not directly comparable to system CPU time
117 measurements. Compare only with other /cpu/classes metrics.
119 /cpu/classes/scavenge/total:cpu-seconds
120 Estimated total CPU time spent performing tasks that return
121 unused memory to the underlying platform. This metric is an
122 overestimate, and not directly comparable to system CPU time
123 measurements. Compare only with other /cpu/classes metrics.
124 Sum of all metrics in /cpu/classes/scavenge.
126 /cpu/classes/total:cpu-seconds
127 Estimated total available CPU time for user Go code or the Go
128 runtime, as defined by GOMAXPROCS. In other words, GOMAXPROCS
129 integrated over the wall-clock duration this process has been
130 executing for. This metric is an overestimate, and not directly
131 comparable to system CPU time measurements. Compare only with
132 other /cpu/classes metrics. Sum of all metrics in /cpu/classes.
134 /cpu/classes/user:cpu-seconds
135 Estimated total CPU time spent running user Go code. This may
136 also include some small amount of time spent in the Go runtime.
137 This metric is an overestimate, and not directly comparable
138 to system CPU time measurements. Compare only with other
139 /cpu/classes metrics.
141 /gc/cycles/automatic:gc-cycles
142 Count of completed GC cycles generated by the Go runtime.
144 /gc/cycles/forced:gc-cycles
145 Count of completed GC cycles forced by the application.
147 /gc/cycles/total:gc-cycles
148 Count of all completed GC cycles.
150 /gc/heap/allocs-by-size:bytes
151 Distribution of heap allocations by approximate size.
152 Note that this does not include tiny objects as defined by
153 /gc/heap/tiny/allocs:objects, only tiny blocks.
155 /gc/heap/allocs:bytes
156 Cumulative sum of memory allocated to the heap by the
159 /gc/heap/allocs:objects
160 Cumulative count of heap allocations triggered by the
161 application. Note that this does not include tiny objects as
162 defined by /gc/heap/tiny/allocs:objects, only tiny blocks.
164 /gc/heap/frees-by-size:bytes
165 Distribution of freed heap allocations by approximate size.
166 Note that this does not include tiny objects as defined by
167 /gc/heap/tiny/allocs:objects, only tiny blocks.
170 Cumulative sum of heap memory freed by the garbage collector.
172 /gc/heap/frees:objects
173 Cumulative count of heap allocations whose storage was freed
174 by the garbage collector. Note that this does not include tiny
175 objects as defined by /gc/heap/tiny/allocs:objects, only tiny
179 Heap size target for the end of the GC cycle.
181 /gc/heap/objects:objects
182 Number of objects, live or unswept, occupying heap memory.
184 /gc/heap/tiny/allocs:objects
185 Count of small allocations that are packed together into blocks.
186 These allocations are counted separately from other allocations
187 because each individual allocation is not tracked by the
188 runtime, only their block. Each block is already accounted for
189 in allocs-by-size and frees-by-size.
191 /gc/limiter/last-enabled:gc-cycle
192 GC cycle the last time the GC CPU limiter was enabled.
193 This metric is useful for diagnosing the root cause of an
194 out-of-memory error, because the limiter trades memory for CPU
195 time when the GC's CPU time gets too high. This is most likely
196 to occur with use of SetMemoryLimit. The first GC cycle is cycle
197 1, so a value of 0 indicates that it was never enabled.
200 Distribution individual GC-related stop-the-world pause
203 /gc/stack/starting-size:bytes
204 The stack size of new goroutines.
206 /godebug/non-default-behavior/execerrdot:events
207 The number of non-default behaviors executed by the os/exec
208 package due to a non-default GODEBUG=execerrdot=... setting.
210 /godebug/non-default-behavior/http2client:events
211 The number of non-default behaviors executed by the net/http
212 package due to a non-default GODEBUG=http2client=... setting.
214 /godebug/non-default-behavior/http2server:events
215 The number of non-default behaviors executed by the net/http
216 package due to a non-default GODEBUG=http2server=... setting.
218 /godebug/non-default-behavior/installgoroot:events
219 The number of non-default behaviors executed by the go/build
220 package due to a non-default GODEBUG=installgoroot=... setting.
222 /godebug/non-default-behavior/multipartfiles:events
223 The number of non-default behaviors executed by
224 the mime/multipart package due to a non-default
225 GODEBUG=multipartfiles=... setting.
227 /godebug/non-default-behavior/multipartmaxheaders:events
228 The number of non-default behaviors executed by
229 the mime/multipart package due to a non-default
230 GODEBUG=multipartmaxheaders=... setting.
232 /godebug/non-default-behavior/multipartmaxparts:events
233 The number of non-default behaviors executed by
234 the mime/multipart package due to a non-default
235 GODEBUG=multipartmaxparts=... setting.
237 /godebug/non-default-behavior/panicnil:events
238 The number of non-default behaviors executed by the runtime
239 package due to a non-default GODEBUG=panicnil=... setting.
241 /godebug/non-default-behavior/randautoseed:events
242 The number of non-default behaviors executed by the math/rand
243 package due to a non-default GODEBUG=randautoseed=... setting.
245 /godebug/non-default-behavior/tarinsecurepath:events
246 The number of non-default behaviors executed by the archive/tar
247 package due to a non-default GODEBUG=tarinsecurepath=...
250 /godebug/non-default-behavior/x509sha1:events
251 The number of non-default behaviors executed by the crypto/x509
252 package due to a non-default GODEBUG=x509sha1=... setting.
254 /godebug/non-default-behavior/x509usefallbackroots:events
255 The number of non-default behaviors executed by the crypto/x509
256 package due to a non-default GODEBUG=x509usefallbackroots=...
259 /godebug/non-default-behavior/zipinsecurepath:events
260 The number of non-default behaviors executed by the archive/zip
261 package due to a non-default GODEBUG=zipinsecurepath=...
264 /memory/classes/heap/free:bytes
265 Memory that is completely free and eligible to be returned to
266 the underlying system, but has not been. This metric is the
267 runtime's estimate of free address space that is backed by
270 /memory/classes/heap/objects:bytes
271 Memory occupied by live objects and dead objects that have not
272 yet been marked free by the garbage collector.
274 /memory/classes/heap/released:bytes
275 Memory that is completely free and has been returned to the
276 underlying system. This metric is the runtime's estimate of free
277 address space that is still mapped into the process, but is not
278 backed by physical memory.
280 /memory/classes/heap/stacks:bytes
281 Memory allocated from the heap that is reserved for stack space,
282 whether or not it is currently in-use.
284 /memory/classes/heap/unused:bytes
285 Memory that is reserved for heap objects but is not currently
286 used to hold heap objects.
288 /memory/classes/metadata/mcache/free:bytes
289 Memory that is reserved for runtime mcache structures, but not
292 /memory/classes/metadata/mcache/inuse:bytes
293 Memory that is occupied by runtime mcache structures that are
294 currently being used.
296 /memory/classes/metadata/mspan/free:bytes
297 Memory that is reserved for runtime mspan structures, but not
300 /memory/classes/metadata/mspan/inuse:bytes
301 Memory that is occupied by runtime mspan structures that are
302 currently being used.
304 /memory/classes/metadata/other:bytes
305 Memory that is reserved for or used to hold runtime metadata.
307 /memory/classes/os-stacks:bytes
308 Stack memory allocated by the underlying operating system.
310 /memory/classes/other:bytes
311 Memory used by execution trace buffers, structures for debugging
312 the runtime, finalizer and profiler specials, and more.
314 /memory/classes/profiling/buckets:bytes
315 Memory that is used by the stack trace hash map used for
318 /memory/classes/total:bytes
319 All memory mapped by the Go runtime into the current process
320 as read-write. Note that this does not include memory mapped
321 by code called via cgo or via the syscall package. Sum of all
322 metrics in /memory/classes.
324 /sched/gomaxprocs:threads
325 The current runtime.GOMAXPROCS setting, or the number of
326 operating system threads that can execute user-level Go code
329 /sched/goroutines:goroutines
330 Count of live goroutines.
332 /sched/latencies:seconds
333 Distribution of the time goroutines have spent in the scheduler
334 in a runnable state before actually running.
336 /sync/mutex/wait/total:seconds
337 Approximate cumulative time goroutines have spent blocked
338 on a sync.Mutex or sync.RWMutex. This metric is useful for
339 identifying global changes in lock contention. Collect a mutex
340 or block profile using the runtime/pprof package for more
341 detailed contention data.