instrument
Module
instrument
Module Summary
Analysis and Utility Functions for Instrumentation
Description
The module instrument contains support for studying the resource usage in an Erlang runtime system. Currently, only the allocation of memory can be studied.
Note that this whole module is experimental, and the representations used as well as the functionality is likely to change in the future.
Data Types
block_histogram() = tuple()A histogram of block sizes where each interval's upper bound is twice as high as the one before it.
The upper bound of the first interval is provided by the function that returned the histogram, and the last interval has no upper bound.
allocation_summary() =
{HistogramStart :: integer() >= 0,
UnscannedSize :: integer() >= 0,
Allocations ::
#{Origin :: atom() =>
#{Type :: atom() => block_histogram()}}}A summary of allocated block sizes (including their headers) grouped by their Origin and Type.
Origin is generally which NIF or driver that allocated the blocks, or 'system' if it could not be determined.
Type is the allocation category that the blocks belong to, e.g. db_term, message or binary.
If one or more carriers could not be scanned in full without harming the responsiveness of the system, UnscannedSize is the number of bytes that had to be skipped.
carrier_info_list() =
{HistogramStart :: integer() >= 0,
Carriers ::
[{AllocatorType :: atom(),
TotalSize :: integer() >= 0,
UnscannedSize :: integer() >= 0,
AllocatedSize :: integer() >= 0,
AllocatedCount :: integer() >= 0,
InPool :: boolean(),
FreeBlocks :: block_histogram()}]}AllocatorType is the type of the allocator that employs this carrier.
TotalSize is the total size of the carrier, including its header.
AllocatedSize is the combined size of the carrier's allocated blocks, including their headers.
AllocatedCount is the number of allocated blocks in the carrier.
InPool is whether the carrier is in the migration pool.
FreeBlocks is a histogram of the free block sizes in the carrier.
If the carrier could not be scanned in full without harming the responsiveness of the system, UnscannedSize is the number of bytes that had to be skipped.
Exports
allocations() -> {ok, Result} | {error, Reason} | OTP 21.0 |
Types
Shorthand for allocations(#{}).
allocations(Options) -> {ok, Result} | {error, Reason} | OTP 21.0 |
Types
Returns a summary of all tagged allocations in the system, optionally filtered by allocator type and scheduler id.
Only binaries and allocations made by NIFs and drivers are tagged by default, but this can be configured an a per-allocator basis with the +M<S>atags emulator option.
If the specified allocator types are not enabled, the call will fail with {error, not_enabled}.
The following options can be used:
allocator_types-
The allocator types that will be searched. Note that blocks can move freely between allocator types, so restricting the search to certain allocators may return unexpected types (e.g. process heaps when searching binary_alloc), or hide blocks that were migrated out.
Defaults to all
alloc_utilallocators. scheduler_ids-
The scheduler ids whose allocator instances will be searched. A scheduler id of 0 will refer to the global instance that is not tied to any particular scheduler. Defaults to all schedulers and the global instance.
histogram_start-
The upper bound of the first interval in the allocated block size histograms. Defaults to 128.
histogram_width-
The number of intervals in the allocated block size histograms. Defaults to 18.
Example:
> instrument:allocations(#{ histogram_start => 128, histogram_width => 15 }).
{ok,{128,0,
#{udp_inet =>
#{driver_event_state => {0,0,0,0,0,0,0,0,0,1,0,0,0,0,0}},
system =>
#{heap => {0,0,0,0,20,4,2,2,2,3,0,1,0,0,1},
db_term => {271,3,1,52,80,1,0,0,0,0,0,0,0,0,0},
code => {0,0,0,5,3,6,11,22,19,20,10,2,1,0,0},
binary => {18,0,0,0,7,0,0,1,0,0,0,0,0,0,0},
message => {0,40,78,2,2,0,0,0,0,0,0,0,0,0,0},
... }
spawn_forker =>
#{driver_select_data_state =>
{1,0,0,0,0,0,0,0,0,0,0,0,0,0,0}},
ram_file_drv => #{drv_binary => {0,0,0,0,0,0,1,0,0,0,0,0,0,0,0}},
prim_file =>
#{process_specific_data => {2,0,0,0,0,0,0,0,0,0,0,0,0,0,0},
nif_trap_export_entry => {0,4,0,0,0,0,0,0,0,0,0,0,0,0,0},
monitor_extended => {0,1,0,0,0,0,0,0,0,0,0,0,0,0,0},
drv_binary => {0,0,0,0,0,0,1,0,3,5,0,0,0,1,0},
binary => {0,4,0,0,0,0,0,0,0,0,0,0,0,0,0}},
prim_buffer =>
#{nif_internal => {0,4,0,0,0,0,0,0,0,0,0,0,0,0,0},
binary => {0,4,0,0,0,0,0,0,0,0,0,0,0,0,0}}}}} carriers() -> {ok, Result} | {error, Reason} | OTP 21.0 |
Types
Shorthand for carriers(#{}).
carriers(Options) -> {ok, Result} | {error, Reason} | OTP 21.0 |
Types
Returns a summary of all carriers in the system, optionally filtered by allocator type and scheduler id.
If the specified allocator types are not enabled, the call will fail with {error, not_enabled}.
The following options can be used:
allocator_types-
The allocator types that will be searched. Defaults to all
alloc_utilallocators. scheduler_ids-
The scheduler ids whose allocator instances will be searched. A scheduler id of 0 will refer to the global instance that is not tied to any particular scheduler. Defaults to all schedulers and the global instance.
histogram_start-
The upper bound of the first interval in the free block size histograms. Defaults to 512.
histogram_width-
The number of intervals in the free block size histograms. Defaults to 14.
Example:
> instrument:carriers(#{ histogram_start => 512, histogram_width => 8 }).
{ok,{512,
[{ll_alloc,1048576,0,1048344,71,false,{0,0,0,0,0,0,0,0}},
{binary_alloc,1048576,0,324640,13,false,{3,0,0,1,0,0,0,2}},
{eheap_alloc,2097152,0,1037200,45,false,{2,1,1,3,4,3,2,2}},
{fix_alloc,32768,0,29544,82,false,{22,0,0,0,0,0,0,0}},
{...}|...]}} See Also
erts_alloc(3), erl(1)
© 2010–2020 Ericsson AB
Licensed under the Apache License, Version 2.0.