MongoDB DBA Associate flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Majority definition

A majority is floor(votingMembers divided by 2) plus 1. A primary must reach a majority of voting members to stay primary; without it, the node steps down and becomes a secondary.

w:majority write concern

The write is acknowledged only after a majority of data-bearing voting members have applied it. With w:1, an unacknowledged-by-majority write can be rolled back if the primary fails immediately after.

electionTimeoutMillis default

Default is 10000 milliseconds. A secondary that has not heard from the primary within this window plus a randomized jitter calls an election. Heartbeats occur every 2 seconds.

Priority 0 member

A replica set member with priority 0 can never become primary. It still replicates data and can serve reads. Hidden members must also have priority 0 and are excluded from read preference selection.

Rollback files

When a node rejoins with oplog entries the new primary never had, MongoDB saves the removed writes to rollback files in the dbPath rollback directory for manual reconciliation.

Chunk size default

Chunks default to 128 MB. The balancer migrates whole chunks between shards, not individual documents, and imbalance is assessed by chunk count per shard.

Hashed versus ranged shard key

A hashed shard key distributes writes evenly but makes range queries scatter-gather. A ranged key supports targeted range reads but risks hotspots on monotonically increasing values.

Targeted versus scatter-gather query

A targeted query includes the full shard key in its filter and routes to one shard. A scatter-gather query omits the shard key and is broadcast to every shard, multiplying cost.

Zone sharding requirement

Each zone must contain at least one shard and each shard belongs to at most one zone. Chunks migrate into a zone only when their shard key range overlaps the zone range.

WiredTiger cache default

The cache defaults to 50 percent of RAM minus 1 GB, or 256 MB if larger. Eviction pressure begins near 80 percent dirty and 95 percent total cache usage.

Journal and j:true

The WiredTiger journal provides crash durability. writeConcern j:true waits for journal acknowledgment; j:false permits loss of writes since the last journal flush, typically up to 100 ms.

BSON document size limit

A single BSON document is capped at 16 MB. Unbounded array growth must be split using patterns such as bucketing or outliers to stay under the limit.

Oplog sizing

The oplog is a capped collection in the local database sized by replication.oplogSizeMB. If a secondary falls behind the oplog window, it must perform a full resync.

Schema validation options

validationLevel is strict or moderate; validationAction is error or warn. Rules are set with $jsonSchema at collection creation or changed later using collMod.

SCRAM-SHA-256

The default authentication mechanism for clients and users. Internal replica set and sharded cluster member authentication instead uses a keyfile or x.509 certificates.

Localhost exception

When authorization is enabled and no users exist, a connection from localhost may create the first user. The exception closes permanently once the first user is created.

Encryption at rest

The WiredTiger encrypted storage engine encrypts data files using a master key held by a key provider such as KMIP or a cloud KMS; the master key is never stored in the data files.

Client-side field level encryption

CSFLE encrypts specified fields in the driver before transmission, so the server stores ciphertext. Queryable encryption additionally supports equality and range queries on the encrypted fields.

authenticationRestrictions

A per-user setting that limits authentication to specified client source IP addresses or CIDR ranges. Connections from outside the allowed ranges are rejected even with valid credentials.

mongodump --oplog

Captures oplog entries that occur during the dump, enabling a point-in-time consistent restore with mongorestore --oplogReplay. It is not supported against a sharded cluster through mongos.

mongorestore --drop

Removes each collection before restoring it. Without --drop, restored documents merge into existing collections, which can produce duplicate documents if the target was not empty.

fsyncLock snapshot procedure

Call db.fsyncLock() to flush and block writes, take the filesystem snapshot, then call db.fsyncUnlock(). Journaling must be enabled for the snapshot to be recoverable.

Point-in-time recovery

Restores a base backup then replays oplog entries up to a target timestamp. The oplog must extend back to the base backup time or the recovery target is unreachable.

Profiler levels

Level 0 is off, level 1 records operations slower than slowms, and level 2 records all operations. Profiling writes to the capped system.profile collection and adds overhead.

explain executionStats fields

totalKeysExamined, totalDocsExamined, and nReturned quantify work. A well-targeted index scan keeps totalDocsExamined close to nReturned, while a large gap signals poor selectivity.

COLLSCAN stage

A collection scan in explain output, meaning no index served the query. It appears as stage COLLSCAN and scales with collection size, unlike an IXSCAN which bounds work by key range.

Replication lag measurement

Compare optimeDate between primary and secondaries in rs.status(). Persistent lag with a small oplog risks the secondary falling off the oplog window and requiring a full resync.

mongotop and mongostat

mongotop reports per-collection read and write time; mongostat reports per-second opcounters, connections, and queue depths. Both sample point-in-time and are best used over intervals.

Rolling upgrade sequence

Upgrade secondaries one at a time, step down the primary so another member takes over, upgrade the former primary, then raise featureCompatibilityVersion as the final step.

featureCompatibilityVersion

FCV gates which features are active and which data formats are permitted. It stays at the previous value after a binary upgrade until setFeatureCompatibilityVersion is run explicitly.

Initial sync

A new replica set member copies all data from a sync source and then applies oplog entries. Large datasets make this slow, so seeding from a filesystem snapshot is often faster.

Ops Manager automation agent

An agent installed on every managed host that performs deployments, restarts, and backup coordination. The backup daemon stores snapshots in a blockstore or object storage bucket.