бэкэнд БД с отображением памяти в slapd (Memory-Mapped DB backend to slapd)
Конфигурация (Configuration)
These slapd.conf options apply to the mdb backend. That is, they
must follow a "backend mdb" line and come before any subsequent
"backend" or "database" lines.
idlexp <exp>
Specify a power of 2 for the maximum size of an index
slot. The default is 16, yielding a maximum slot size of
2^16 or 65536. Once set, this option applies to every mdb
database instance. The specified value must be in the
range of 16-30.
These slapd.conf options apply to the mdb backend database. That
is, they must follow a "database mdb" line and come before any
subsequent "backend" or "database" lines. Other database options
are described in the slapd.conf(5) manual page.
checkpoint <kbyte> <min>
Specify the frequency for flushing the database disk
buffers. This setting is only needed if the dbnosync
option is used. The checkpoint will occur if either
<kbyte> data has been written or <min> minutes have passed
since the last checkpoint. Both arguments default to
zero, in which case they are ignored. When the <min>
argument is non-zero, an internal task will run every
<min> minutes to perform the checkpoint. Note: currently
the <kbyte> setting is unimplemented.
dbnosync
Specify that on-disk database contents should not be
immediately synchronized with in memory changes. Enabling
this option may improve performance at the expense of data
security. In particular, if the operating system crashes
before changes are flushed, some number of transactions
may be lost. By default, a full data flush/sync is
performed when each transaction is committed.
directory <directory>
Specify the directory where the LMDB files containing this
database and associated indexes live. A separate
directory must be specified for each database. The
default is LOCALSTATEDIR/openldap-data.
envflags {nosync,nometasync,writemap,mapasync,nordahead}
Specify flags for finer-grained control of the LMDB
library's operation.
nosync This is exactly the same as the dbnosync directive.
nometasync
Flush the data on a commit, but skip the sync of
the meta page. This mode is slightly faster than
doing a full sync, but can potentially lose the
last committed transaction if the operating system
crashes. If both nometasync and nosync are set, the
nosync flag takes precedence.
writemap
Use a writable memory map instead of just read-
only. This speeds up write operations but makes the
database vulnerable to corruption in case any bugs
in slapd cause stray writes into the mmap region.
mapasync
When using a writable memory map and performing
flushes on each commit, use an asynchronous flush
instead of a synchronous flush (the default). This
option has no effect if writemap has not been set.
It also has no effect if nosync is set.
nordahead
Turn off file readahead. Usually the OS performs
readahead on every read request. This usually
boosts read performance but can be harmful to
random access read performance if the system's
memory is full and the DB is larger than RAM. This
option is not implemented on Windows.
index {<attrlist>|default} [pres,eq,approx,sub,<special>]
Specify the indexes to maintain for the given attribute
(or list of attributes). Some attributes only support a
subset of indexes. If only an <attr> is given, the
indices specified for default are maintained. Note that
setting a default does not imply that all attributes will
be indexed. Also, for best performance, an eq index should
always be configured for the objectClass attribute.
A number of special index parameters may be specified.
The index type sub can be decomposed into subinitial,
subany, and subfinal indices. The special type nolang may
be specified to disallow use of this index by language
subtypes. The special type nosubtypes may be specified to
disallow use of this index by named subtypes. Note:
changing index settings in slapd.conf(5) requires
rebuilding indices, see slapindex(8); changing index
settings dynamically by LDAPModifying "cn=config"
automatically causes rebuilding of the indices online in a
background task.
maxentrysize <bytes>
Specify the maximum size of an entry in bytes. Attempts to
store an entry larger than this size will be rejected with
the error LDAP_ADMINLIMIT_EXCEEDED. The default is 0,
which is unlimited.
maxreaders <integer>
Specify the maximum number of threads that may have
concurrent read access to the database. Tools such as
slapcat count as a single thread, in addition to threads
in any active slapd processes. The default is 126.
maxsize <bytes>
Specify the maximum size of the database in bytes. A
memory map of this size is allocated at startup time and
the database will not be allowed to grow beyond this size.
The default is 10485760 bytes. This setting may be changed
upward if the configured limit needs to be increased.
Note: It is important to set this to as large a value as
possible, (relative to anticipated growth of the actual
data over time) since growing the size later may not be
practical when the system is under heavy load.
mode <integer>
Specify the file protection mode that newly created
database files should have. The default is 0600.
multival {<attrlist>|default} <integer hi>,<integer lo>
Specify the number of values for which a multivalued
attribute is stored in a separate table. Normally entries
are stored as a single blob inside the database. When an
entry gets very large or contains attributes with a very
large number of values, modifications on that entry may
get very slow. Splitting the large attributes out to a
separate table can improve the performance of modification
operations. The threshold is specified as a pair of
integers. If the number of values exceeds the hi threshold
the values will be split out. If a modification deletes
enough values to bring an attribute below the lo threshold
the values will be removed from the separate table and
merged back into the main entry blob. The threshold can
be set for a specific list of attributes, or the default
can be configured for all other attributes. The default
value for both hi and lo thresholds is UINT_MAX, which
keeps all attributes in the main blob.
rtxnsize <entries>
Specify the maximum number of entries to process in a
single read transaction when executing a large search.
Long-lived read transactions prevent old database pages
from being reused in write transactions, and so can cause
significant growth of the database file when there is
heavy write traffic. This setting causes the read
transaction in large searches to be released and
reacquired after the given number of entries has been
read, to give writers the opportunity to reclaim old
database pages. The default is 10000.
searchstack <depth>
Specify the depth of the stack used for search filter
evaluation. Search filters are evaluated on a stack to
accommodate nested AND / OR clauses. An individual stack
is assigned to each server thread. The depth of the stack
determines how complex a filter can be evaluated without
requiring any additional memory allocation. Filters that
are nested deeper than the search stack depth will cause a
separate stack to be allocated for that particular search
operation. These allocations can have a major negative
impact on server performance, but specifying too much
stack will also consume a great deal of memory. Each
search stack uses 512K bytes per level. The default stack
depth is 16, thus 8MB per thread is used.
