OPS: Ops Jumpstart: Admin 101
JSON
Schema-less
More agile without schema
Db.persons.find ( {“age”: {“$gt”: 40}})
Db.persons.update( {“name”: “James Bond”}, { “$inc” : “age”})
No pretense of looking like SQL, but very functional
TB, PB of data
Built for huge
Replica Set
Primary server
Secondary server
Secondary server
Replicas of data.
All writes go to primary first, broadcast to secondarys, kept in sync
Rolling social networks
Not for bigger or scaling
Sharding
Multiple databases in parallel (shards)
Each shard has part of the data
Shard 0 (a-c)
Shard 1 (d-h)
Mongo Router (mongos) finds the shards
Shards can’t get down, must be always available
Can’t lose a shard
Combine sharding and replication
Each shard is a replica set
Replica set (3,5,7,…)
Smaller servers, small license fee
Dealing with dozens and hundreds of servers is a pain
Real analytics
Document store
Query language
scalability
Replication (HA)
Sharding (volume, throughput)
No schemas
Indexes are extremely important
Security is a big deal
Lots of servers
Monitoring
Backups
Deployment
Every mongo query sould have an index, unless less then 1000 docs in collection
More than 50% of issues is index
Find, define, use
Database without a password, important for local use
Small mongodb per application, easy to tune
MongoDB MMS
Performance data to mongo cloud
Network dial out to mongodb for stats
2 factor auth with password
Easy for support
Enterprise mongo can be installed and run locally
Backup, deploy
MMS (in cloud)
Works with tech support 20-30k customers using
Ops Manager (on prem)
Run yourself in your network
Mms on prem
Hidden replica for backup
Snapshot every six hours plus oplog
Point-in-time restore
3-4x original storage for typical retention period
100s of gb/day of data over net
On-prem for enterprise
Backup in the cloud – automatic don’t need space
The post MongoDB World – June 1, 2015 appeared first on MongoDB Developer.