In-Memory Storage Engine
On this page
Changed in version 3.2.6.
Starting in MongoDB Enterprise version 3.2.6, the in-memory storage engine is part of general availability (GA) in the 64-bit builds. Other than some metadata and diagnostic data, the in-memory storage engine does not maintain any on-disk data, including configuration data, indexes, user credentials, etc.
By avoiding disk I/O, the in-memory storage engine allows for more predictable latency of database operations.
Specify In-Memory Storage Engine
To select the in-memory storage engine, specify:
inMemory
for the--storageEngine
option, or thestorage.engine
setting if using a configuration file.--dbpath
, orstorage.dbPath
if using a configuration file. Although the in-memory storage engine does not write data to the filesystem, it maintains in the--dbpath
small metadata files and diagnostic data as well temporary files for building large indexes.
For example, from the command line:
mongod --storageEngine inMemory --dbpath <path>
Or, if using the YAML configuration file format:
storage: engine: inMemory dbPath: <path>
See inMemory Options for configuration options specific to
this storage engine. Most mongod
configuration options are
available for use with in-memory storage engine except for those
options that are related to data persistence, such as journaling or
encryption at rest configuration.
Warning
The in-memory storage engine does not persist data after process shutdown.
Concurrency
The in-memory storage engine uses document-level concurrency control for write operations. As a result, multiple clients can modify different documents of a collection at the same time.
Memory Use
In-memory storage engine requires that all its data (including indexes,
oplog if mongod
instance is part of a replica set, etc.) must
fit into the specified --inMemorySizeGB
command-line option
or storage.inMemory.engineConfig.inMemorySizeGB
setting in
the YAML configuration file.
By default, the in-memory storage engine uses 50% of physical RAM minus 1 GB.
If a write operation would cause the data to exceed the specified memory size, MongoDB returns with the error:
"WT_CACHE_FULL: operation would overflow cache"
To specify a new size, use the
storage.inMemory.engineConfig.inMemorySizeGB
setting in the
YAML configuration file format:
storage: engine: inMemory dbPath: <path> inMemory: engineConfig: inMemorySizeGB: <newSize>
Or use the command-line option --inMemorySizeGB
:
mongod --storageEngine inMemory --dbpath <path> --inMemorySizeGB <newSize>
Durability
The in-memory storage engine is non-persistent and does not write data to a persistent storage. Non-persisted data includes application data and system data, such as users, permissions, indexes, replica set configuration, sharded cluster configuration, etc.
As such, the concept of journal or waiting for data to become durable does not apply to the in-memory storage engine.
If any voting member of a replica set uses the in-memory
storage engine, you must set
writeConcernMajorityJournalDefault
to false
.
Note
Starting in version 4.2 (and 4.0.13 and 3.6.14 ), if a replica set
member uses the in-memory storage engine
(voting or non-voting) but the replica set has
writeConcernMajorityJournalDefault
set to true, the
replica set member logs a startup warning.
With writeConcernMajorityJournalDefault
set to false
,
MongoDB does not wait for w: "majority"
writes to be written to the on-disk journal before acknowledging the
writes. As such, "majority"
write operations could
possibly roll back in the event of a transient loss (e.g. crash and
restart) of a majority of nodes in a given replica set.
Write operations that specify a write concern journaled
are acknowledged immediately. When an mongod
instance
shuts down, either as result of the shutdown
command or
due to a system error, recovery of in-memory data is impossible.
Transactions
Transactions are supported on replica sets and sharded clusters where:
the primary uses the WiredTiger storage engine, and
the secondary members use either the WiredTiger storage engine or the in-memory storage engines.
Note
You cannot run transactions on a sharded cluster that has a shard
with writeConcernMajorityJournalDefault
set to
false
, such as a shard with a voting member that uses the
in-memory storage engine.
Deployment Architectures
In addition to running as standalones, mongod
instances that
use in-memory storage engine can run as part of a replica set or part
of a sharded cluster.
Replica Set
You can deploy mongod
instances that use in-memory storage
engine as part of a replica set. For example, as part of a three-member
replica set, you could have:
two
mongod
instances run with in-memory storage engine.one
mongod
instance run with WiredTiger storage engine. Configure the WiredTiger member as a hidden member (i.e.hidden: true
andpriority: 0
).
With this deployment model, only the mongod
instances
running with the in-memory storage engine can become the primary.
Clients connect only to the in-memory storage engine mongod
instances. Even if both mongod
instances running in-memory
storage engine crash and restart, they can sync from the member running
WiredTiger. The hidden mongod
instance running with
WiredTiger persists the data to disk, including the user data, indexes,
and replication configuration information.
Note
In-memory storage engine requires that all its data (including oplog
if mongod
is part of replica set, etc.) fit into the
specified --inMemorySizeGB
command-line option or
storage.inMemory.engineConfig.inMemorySizeGB
setting. See
Memory Use.
Sharded Cluster
You can deploy mongod
instances that use an in-memory
storage engine as part of a sharded cluster. The in-memory
storage engine avoids disk I/O to allow for more
predictable database operation latency. In a sharded cluster, a
shard can consist of a single mongod
instance or a
replica set. For example, you could have one shard that
consists of the following replica set:
two
mongod
instances run with in-memory storage engineone
mongod
instance run with WiredTiger storage engine. Configure the WiredTiger member as a hidden member (i.e.hidden: true
andpriority: 0
).
To this shard, add the tag
inmem
. For
example, if this shard has the name shardC
, connect to the
mongos
and run sh.addShardTag()
.
For example,
sh.addShardTag("shardC", "inmem")
To the other shards, add a separate tag persisted
.
sh.addShardTag("shardA", "persisted") sh.addShardTag("shardB", "persisted")
For each sharded collection that should reside on the inmem
shard,
assign to the entire chunk range
the tag
inmem
:
sh.addTagRange("test.analytics", { shardKey: MinKey }, { shardKey: MaxKey }, "inmem")
For each sharded collection that should reside across the
persisted
shards, assign to the entire chunk range
the tag
persisted
:
sh.addTagRange("salesdb.orders", { shardKey: MinKey }, { shardKey: MaxKey }, "persisted")
For the inmem
shard, create a database or move the database.
Note
Read concern level "snapshot"
is not officially supported
with the in-memory storage engine.