Source file src/runtime/metrics/doc.go

     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.
     4  
     5  /*
     6  Package metrics provides a stable interface to access implementation-defined
     7  metrics exported by the Go runtime. This package is similar to existing functions
     8  like runtime.ReadMemStats and debug.ReadGCStats, but significantly more general.
     9  
    10  The set of metrics defined by this package may evolve as the runtime itself
    11  evolves, and also enables variation across Go implementations, whose relevant
    12  metric sets may not intersect.
    13  
    14  # Interface
    15  
    16  Metrics are designated by a string key, rather than, for example, a field name in
    17  a struct. The full list of supported metrics is always available in the slice of
    18  Descriptions returned by All. Each Description also includes useful information
    19  about the metric.
    20  
    21  Thus, users of this API are encouraged to sample supported metrics defined by the
    22  slice returned by All to remain compatible across Go versions. Of course, situations
    23  arise where reading specific metrics is critical. For these cases, users are
    24  encouraged to use build tags, and although metrics may be deprecated and removed,
    25  users should consider this to be an exceptional and rare event, coinciding with a
    26  very large change in a particular Go implementation.
    27  
    28  Each metric key also has a "kind" that describes the format of the metric's value.
    29  In the interest of not breaking users of this package, the "kind" for a given metric
    30  is guaranteed not to change. If it must change, then a new metric will be introduced
    31  with a new key and a new "kind."
    32  
    33  # Metric key format
    34  
    35  As mentioned earlier, metric keys are strings. Their format is simple and well-defined,
    36  designed to be both human and machine readable. It is split into two components,
    37  separated by a colon: a rooted path and a unit. The choice to include the unit in
    38  the key is motivated by compatibility: if a metric's unit changes, its semantics likely
    39  did also, and a new key should be introduced.
    40  
    41  For more details on the precise definition of the metric key's path and unit formats, see
    42  the documentation of the Name field of the Description struct.
    43  
    44  # A note about floats
    45  
    46  This package supports metrics whose values have a floating-point representation. In
    47  order to improve ease-of-use, this package promises to never produce the following
    48  classes of floating-point values: NaN, infinity.
    49  
    50  # Supported metrics
    51  
    52  Below is the full list of supported metrics, ordered lexicographically.
    53  
    54  	/cgo/go-to-c-calls:calls
    55  		Count of calls made from Go to C by the current process.
    56  
    57  	/cpu/classes/gc/mark/assist:cpu-seconds
    58  		Estimated total CPU time goroutines spent performing GC tasks
    59  		to assist the GC and prevent it from falling behind the application.
    60  		This metric is an overestimate, and not directly comparable to
    61  		system CPU time measurements. Compare only with other /cpu/classes
    62  		metrics.
    63  
    64  	/cpu/classes/gc/mark/dedicated:cpu-seconds
    65  		Estimated total CPU time spent performing GC tasks on
    66  		processors (as defined by GOMAXPROCS) dedicated to those tasks.
    67  		This includes time spent with the world stopped due to the GC.
    68  		This metric is an overestimate, and not directly comparable to
    69  		system CPU time measurements. Compare only with other /cpu/classes
    70  		metrics.
    71  
    72  	/cpu/classes/gc/mark/idle:cpu-seconds
    73  		Estimated total CPU time spent performing GC tasks on
    74  		spare CPU resources that the Go scheduler could not otherwise find
    75  		a use for. This should be subtracted from the total GC CPU time to
    76  		obtain a measure of compulsory GC CPU time.
    77  		This metric is an overestimate, and not directly comparable to
    78  		system CPU time measurements. Compare only with other /cpu/classes
    79  		metrics.
    80  
    81  	/cpu/classes/gc/pause:cpu-seconds
    82  		Estimated total CPU time spent with the application paused by
    83  		the GC. Even if only one thread is running during the pause, this is
    84  		computed as GOMAXPROCS times the pause latency because nothing else
    85  		can be executing. This is the exact sum of samples in /gc/pause:seconds
    86  		if each sample is multiplied by GOMAXPROCS at the time it is taken.
    87  		This metric is an overestimate, and not directly comparable to
    88  		system CPU time measurements. Compare only with other /cpu/classes
    89  		metrics.
    90  
    91  	/cpu/classes/gc/total:cpu-seconds
    92  		Estimated total CPU time spent performing GC tasks.
    93  		This metric is an overestimate, and not directly comparable to
    94  		system CPU time measurements. Compare only with other /cpu/classes
    95  		metrics. Sum of all metrics in /cpu/classes/gc.
    96  
    97  	/cpu/classes/idle:cpu-seconds
    98  		Estimated total available CPU time not spent executing any Go or Go
    99  		runtime code. In other words, the part of /cpu/classes/total:cpu-seconds
   100  		that was unused.
   101  		This metric is an overestimate, and not directly comparable to
   102  		system CPU time measurements. Compare only with other /cpu/classes
   103  		metrics.
   104  
   105  	/cpu/classes/scavenge/assist:cpu-seconds
   106  		Estimated total CPU time spent returning unused memory to the
   107  		underlying platform in response eagerly in response to memory pressure.
   108  		This metric is an overestimate, and not directly comparable to
   109  		system CPU time measurements. Compare only with other /cpu/classes
   110  		metrics.
   111  
   112  	/cpu/classes/scavenge/background:cpu-seconds
   113  		Estimated total CPU time spent performing background tasks
   114  		to return unused memory to the underlying platform.
   115  		This metric is an overestimate, and not directly comparable to
   116  		system CPU time measurements. Compare only with other /cpu/classes
   117  		metrics.
   118  
   119  	/cpu/classes/scavenge/total:cpu-seconds
   120  		Estimated total CPU time spent performing tasks that return
   121  		unused memory to the underlying platform.
   122  		This metric is an overestimate, and not directly comparable to
   123  		system CPU time measurements. Compare only with other /cpu/classes
   124  		metrics. Sum of all metrics in /cpu/classes/scavenge.
   125  
   126  	/cpu/classes/total:cpu-seconds
   127  		Estimated total available CPU time for user Go code or the Go runtime, as
   128  		defined by GOMAXPROCS. In other words, GOMAXPROCS integrated over the
   129  		wall-clock duration this process has been executing for.
   130  		This metric is an overestimate, and not directly comparable to
   131  		system CPU time measurements. Compare only with other /cpu/classes
   132  		metrics. Sum of all metrics in /cpu/classes.
   133  
   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 to
   138  		system CPU time measurements. Compare only with other /cpu/classes
   139  		metrics.
   140  
   141  	/gc/cycles/automatic:gc-cycles
   142  		Count of completed GC cycles generated by the Go runtime.
   143  
   144  	/gc/cycles/forced:gc-cycles
   145  		Count of completed GC cycles forced by the application.
   146  
   147  	/gc/cycles/total:gc-cycles
   148  		Count of all completed GC cycles.
   149  
   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 /gc/heap/tiny/allocs:objects,
   153  		only tiny blocks.
   154  
   155  	/gc/heap/allocs:bytes
   156  		Cumulative sum of memory allocated to the heap by the application.
   157  
   158  	/gc/heap/allocs:objects
   159  		Cumulative count of heap allocations triggered by the application.
   160  		Note that this does not include tiny objects as defined by /gc/heap/tiny/allocs:objects,
   161  		only tiny blocks.
   162  
   163  	/gc/heap/frees-by-size:bytes
   164  		Distribution of freed heap allocations by approximate size.
   165  		Note that this does not include tiny objects as defined by /gc/heap/tiny/allocs:objects,
   166  		only tiny blocks.
   167  
   168  	/gc/heap/frees:bytes
   169  		Cumulative sum of heap memory freed by the garbage collector.
   170  
   171  	/gc/heap/frees:objects
   172  		Cumulative count of heap allocations whose storage was freed by the garbage collector.
   173  		Note that this does not include tiny objects as defined by /gc/heap/tiny/allocs:objects,
   174  		only tiny blocks.
   175  
   176  	/gc/heap/goal:bytes
   177  		Heap size target for the end of the GC cycle.
   178  
   179  	/gc/heap/objects:objects
   180  		Number of objects, live or unswept, occupying heap memory.
   181  
   182  	/gc/heap/tiny/allocs:objects
   183  		Count of small allocations that are packed together into blocks.
   184  		These allocations are counted separately from other allocations
   185  		because each individual allocation is not tracked by the runtime,
   186  		only their block. Each block is already accounted for in
   187  		allocs-by-size and frees-by-size.
   188  
   189  	/gc/limiter/last-enabled:gc-cycle
   190  		GC cycle the last time the GC CPU limiter was enabled.
   191  		This metric is useful for diagnosing the root cause of an out-of-memory
   192  		error, because the limiter trades memory for CPU time when the GC's CPU
   193  		time gets too high. This is most likely to occur with use of SetMemoryLimit.
   194  		The first GC cycle is cycle 1, so a value of 0 indicates that it was never enabled.
   195  
   196  	/gc/pauses:seconds
   197  		Distribution individual GC-related stop-the-world pause latencies.
   198  
   199  	/gc/stack/starting-size:bytes
   200  		The stack size of new goroutines.
   201  
   202  	/memory/classes/heap/free:bytes
   203  		Memory that is completely free and eligible to be returned to
   204  		the underlying system, but has not been. This metric is the
   205  		runtime's estimate of free address space that is backed by
   206  		physical memory.
   207  
   208  	/memory/classes/heap/objects:bytes
   209  		Memory occupied by live objects and dead objects that have
   210  		not yet been marked free by the garbage collector.
   211  
   212  	/memory/classes/heap/released:bytes
   213  		Memory that is completely free and has been returned to
   214  		the underlying system. This metric is the runtime's estimate of
   215  		free address space that is still mapped into the process, but
   216  		is not backed by physical memory.
   217  
   218  	/memory/classes/heap/stacks:bytes
   219  		Memory allocated from the heap that is reserved for stack
   220  		space, whether or not it is currently in-use.
   221  
   222  	/memory/classes/heap/unused:bytes
   223  		Memory that is reserved for heap objects but is not currently
   224  		used to hold heap objects.
   225  
   226  	/memory/classes/metadata/mcache/free:bytes
   227  		Memory that is reserved for runtime mcache structures, but
   228  		not in-use.
   229  
   230  	/memory/classes/metadata/mcache/inuse:bytes
   231  		Memory that is occupied by runtime mcache structures that
   232  		are currently being used.
   233  
   234  	/memory/classes/metadata/mspan/free:bytes
   235  		Memory that is reserved for runtime mspan structures, but
   236  		not in-use.
   237  
   238  	/memory/classes/metadata/mspan/inuse:bytes
   239  		Memory that is occupied by runtime mspan structures that are
   240  		currently being used.
   241  
   242  	/memory/classes/metadata/other:bytes
   243  		Memory that is reserved for or used to hold runtime
   244  		metadata.
   245  
   246  	/memory/classes/os-stacks:bytes
   247  		Stack memory allocated by the underlying operating system.
   248  
   249  	/memory/classes/other:bytes
   250  		Memory used by execution trace buffers, structures for
   251  		debugging the runtime, finalizer and profiler specials, and
   252  		more.
   253  
   254  	/memory/classes/profiling/buckets:bytes
   255  		Memory that is used by the stack trace hash map used for
   256  		profiling.
   257  
   258  	/memory/classes/total:bytes
   259  		All memory mapped by the Go runtime into the current process
   260  		as read-write. Note that this does not include memory mapped
   261  		by code called via cgo or via the syscall package.
   262  		Sum of all metrics in /memory/classes.
   263  
   264  	/sched/gomaxprocs:threads
   265  		The current runtime.GOMAXPROCS setting, or the number of
   266  		operating system threads that can execute user-level Go code
   267  		simultaneously.
   268  
   269  	/sched/goroutines:goroutines
   270  		Count of live goroutines.
   271  
   272  	/sched/latencies:seconds
   273  		Distribution of the time goroutines have spent in the scheduler
   274  		in a runnable state before actually running.
   275  
   276  	/sync/mutex/wait/total:seconds
   277  		Approximate cumulative time goroutines have spent blocked on a
   278  		sync.Mutex or sync.RWMutex. This metric is useful for identifying
   279  		global changes in lock contention. Collect a mutex or block
   280  		profile using the runtime/pprof package for more detailed
   281  		contention data.
   282  */
   283  package metrics
   284  

View as plain text