| PMEMOBJ_CTL_GET(3) | PMDK Programmer's Manual | PMEMOBJ_CTL_GET(3) |
pmemobj_ctl_get(), pmemobj_ctl_set(), pmemobj_ctl_exec() - Query and modify libpmemobj internal behavior (EXPERIMENTAL)
#include <libpmemobj.h> int pmemobj_ctl_get(PMEMobjpool *pop, const char *name, void *arg); (EXPERIMENTAL) int pmemobj_ctl_set(PMEMobjpool *pop, const char *name, void *arg); (EXPERIMENTAL) int pmemobj_ctl_exec(PMEMobjpool *pop, const char *name, void *arg); (EXPERIMENTAL)
The pmemobj_ctl_get(), pmemobj_ctl_set() and pmemobj_ctl_exec() functions provide a uniform interface for querying and modifying the internal behavior of libpmemobj(7) through the control (CTL) namespace.
The name argument specifies an entry point as defined in the CTL namespace specification. The entry point description specifies whether the extra arg is required. Those two parameters together create a CTL query. The functions and the entry points are thread-safe unless indicated otherwise below. If there are special conditions for calling an entry point, they are explicitly stated in its description. The functions propagate the return value of the entry point. If either name or arg is invalid, -1 is returned.
If the provided ctl query is valid, the CTL functions will always return 0 on success and -1 on failure, unless otherwise specified in the entry point description.
See more in pmem_ctl(5) man page.
prefault.at_create | rw | global | int | int | - | boolean
If set, every page of the pool will be touched and written to when the pool is created, in order to trigger page allocation and minimize the performance impact of pagefaults. Affects only the pmemobj_create() function.
prefault.at_open | rw | global | int | int | - | boolean
If set, every page of the pool will be touched and written to when the pool is opened, in order to trigger page allocation and minimize the performance impact of pagefaults. Affects only the pmemobj_open() function.
sds.at_create | rw | global | int | int | - | boolean
If set, force-enables or force-disables SDS feature during pool creation. Affects only the pmemobj_create() function. See pmempool_feature_query(3) for informations about SDS (SHUTDOWN_STATE) feature.
tx.debug.skip_expensive_checks | rw | - | int | int | - | boolean
Turns off some expensive checks performed by the transaction module in “debug” builds. Ignored in “release” builds.
tx.cache.size | rw | - | long long | long long | - | integer
Size in bytes of the transaction snapshot cache. In a larger cache the frequency of persistent allocations is lower, but with higher fixed cost.
This should be set to roughly the sum of sizes of the snapshotted regions in an average transaction in the pool.
This entry point is not thread safe and should not be modified if there are any transactions currently running.
This value must be a in a range between 0 and PMEMOBJ_MAX_ALLOC_SIZE, otherwise this entry point will fail.
tx.cache.threshold | rw | - | long long | long long | - | integer
This entry point is deprecated. All snapshots, regardless of the size, use the transactional cache.
tx.post_commit.queue_depth | rw | - | int | int | - | integer
This entry point is deprecated.
tx.post_commit.worker | r- | - | void * | - | - | -
This entry point is deprecated.
tx.post_commit.stop | r- | - | void * | - | - | -
This entry point is deprecated.
heap.alloc_class.[class_id].desc | rw | - | struct pobj_alloc_class_desc | struct pobj_alloc_class_desc | - | integer, integer, integer, string
Describes an allocation class. Allows one to create or view the internal data structures of the allocator.
Creating custom allocation classes can be beneficial for both raw allocation throughput, scalability and, most importantly, fragmentation. By carefully constructing allocation classes that match the application workload, one can entirely eliminate external and internal fragmentation. For example, it is possible to easily construct a slab-like allocation mechanism for any data structure.
The [class_id] is an index field. Only values between 0-254 are valid. If setting an allocation class, but the class_id is already taken, the function will return -1. The values between 0-127 are reserved for the default allocation classes of the library and can be used only for reading.
The recommended method for retrieving information about all allocation classes is to call this entry point for all class ids between 0 and 254 and discard those results for which the function returns an error.
This entry point takes a complex argument.
struct pobj_alloc_class_desc {
size_t unit_size;
size_t alignment;
unsigned units_per_block;
enum pobj_header_type header_type;
unsigned class_id;
};
The first field, unit_size, is an 8-byte unsigned integer that defines the allocation class size. While theoretically limited only by PMEMOBJ_MAX_ALLOC_SIZE, for most workloads this value should be between 8 bytes and 2 megabytes.
The alignment field specifies the user data alignment of objects allocated using the class. If set, must be a power of two and an even divisor of unit size. Alignment is limited to maximum of 2 megabytes. All objects have default alignment of 64 bytes, but the user data alignment is affected by the size of the chosen header.
The units_per_block field defines how many units a single block of memory contains. This value will be adjusted to match the internal size of the block (256 kilobytes or a multiple thereof). For example, given a class with a unit_size of 512 bytes and a units_per_block of 1000, a single block of memory for that class will have 512 kilobytes. This is relevant because the bigger the block size, the less frequently blocks need to be fetched, resulting in lower contention on global heap state. If the CTL call is being done at runtime, the units_per_block variable of the provided alloc class structure is modified to match the actual value.
The header_type field defines the header of objects from the allocation class. There are three types:
The class_id field is an optional, runtime-only variable that allows the user to retrieve the identifier of the class. This will be equivalent to the provided [class_id]. This field cannot be set from a config file.
The allocation classes are a runtime state of the library and must be created after every open. It is highly recommended to use the configuration file to store the classes.
This structure is declared in the libpmemobj/ctl.h header file. Please refer to this file for an in-depth explanation of the allocation classes and relevant algorithms.
Allocation classes constructed in this way can be leveraged by explicitly specifying the class using POBJ_CLASS_ID(id) flag in pmemobj_tx_xalloc()/pmemobj_xalloc() functions.
Example of a valid alloc class query string:
heap.alloc_class.128.desc=500,0,1000,compact
This query, if executed, will create an allocation class with an id of 128 that has a unit size of 500 bytes, has at least 1000 units per block and uses a compact header.
For reading, function returns 0 if successful, if the allocation class does not exist it sets the errno to ENOENT and returns -1;
This entry point can fail if any of the parameters of the allocation class is invalid or if exactly the same class already exists.
heap.alloc_class.new.desc | -w | - | - | struct pobj_alloc_class_desc | - | integer, integer, integer, string
Same as heap.alloc_class.[class_id].desc, but instead of requiring the user to provide the class_id, it automatically creates the allocation class with the first available identifier.
This should be used when it's impossible to guarantee unique allocation class naming in the application (e.g. when writing a library that uses libpmemobj).
The required class identifier will be stored in the class_id field of the struct pobj_alloc_class_desc.
stats.enabled | rw | - | int | int | - | boolean
Enables or disables runtime collection of statistics. Statistics are not recalculated after enabling; any operations that occur between disabling and re-enabling will not be reflected in subsequent values.
Statistics are disabled by default. Enabling them may have non-trivial performance impact.
stats.heap.curr_allocated | r- | - | int | - | - | -
Reads the number of bytes currently allocated in the heap. If statistics were disabled at any time in the lifetime of the heap, this value may be inaccurate.
heap.size.granularity | rw- | - | uint64_t | uint64_t | - | long long
Reads or modifies the granularity with which the heap grows when OOM. Valid only if the poolset has been defined with directories.
A granularity of 0 specifies that the pool will not grow automatically.
This entry point can fail if the granularity value is non-zero and smaller than PMEMOBJ_MIN_PART.
heap.size.extend | –x | - | - | - | uint64_t | -
Extends the heap by the given size. Must be larger than PMEMOBJ_MIN_PART.
This entry point can fail if the pool does not support extend functionality or if there's not enough space left on the device.
debug.heap.alloc_pattern | rw | - | int | int | - | -
Single byte pattern that is used to fill new uninitialized memory allocation. If the value is negative, no pattern is written. This is intended for debugging, and is disabled by default.
In addition to direct function call, each write entry point can also be set using two alternative methods.
The first method is to load a configuration directly from the PMEMOBJ_CONF environment variable.
The second method of loading an external configuration is to set the PMEMOBJ_CONF_FILE environment variable to point to a file that contains a sequence of ctl queries.
See more in pmem_ctl(5) man page.
libpmemobj(7), pmem_ctl(5) and <http://pmem.io>
| 2019-02-19 | PMDK - pmemobj API version 2.3 |