<\/p>\n
Yesterday in one of our mongodb databases developer asked for unshard a sharded collection.<\/p>\n
Yes it is possible to unshard a sharded collection but it is not a cute process \ud83d\ude42 \u00a0\u00a0As the data has propagated between 3 servers, at first we need to take all data into primary server.<\/p>\n
In addition to this, this process needs down time.<\/p>\n
Before moving data to the primary shard we need to shutdown all shards and all mongos except one.<\/p>\n
After finishing the changes at that mongos instance, wee need to restart the mongos, too.<\/p>\n
database = 'db'\r\n\r\ncollection = database + '.fs.chunks'\r\n\r\n\u00a0sh.stopBalancer()\r\n\r\nuse config\r\n\r\nprimary = db.databases.findOne({_id: database}).primary\r\n\r\n\u00a0\/\/ move all chunks to primary\r\n\r\ndb.chunks.find({ns: collection, shard: {$ne: primary}}).forEach(function(chunk){print('moving chunk from', chunk.shard, 'to', primary, '::', tojson(chunk.min), '-->', tojson(chunk.max));sh.moveChunk(collection, chunk.min, primary);});\r\n\r\n\u00a0\/\/ unshard\r\n\r\ndb.collections.remove({ \"_id\" : \"test.fs.chunks\" })\r\n\r\ndb.chunks.remove({ ns : collection })<\/pre>\n <\/p>\n
Here we can start shards and other mongos.<\/p>\n
\/\/ flush all mongos\r\n\r\nuse admin\r\n\r\ndb.runCommand({ flushRouterConfig: 1 })<\/pre>\n <\/p>\n
As this process has many manual steps it is a little challenging process.<\/p>\n
The size of the collection to make unsharded is not\u00a0 big so we used another method.<\/p>\n
First we create a copy collection by insert.<\/p>\n
db.Collection.find({}).forEach(\u00a0\r\n\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0function(myDoc) {\u00a0\r\n\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0db.CollectionCopy.insert(myDoc);\r\n\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0}\u00a0\r\n\r\n);<\/pre>\n <\/p>\n
After the insert process finishes, we dropped the collection named ‘Collection’.<\/p>\n
While Dropping a sharded collection, it is important to validate the dropping process is succeeded in all the shards and in config database.<\/p>\n
Dropping a Collection in a Sharded Cluster<\/p>\n
1- Drop the collection using a mongos<\/p>\n
use db\r\n\r\ndb.Collection.drop()<\/pre>\n2- Connect to each shard’s primary and verify the namespace has been dropped. If it has not, please drop it.<\/p>\n
3- Connect to a mongos, switch to the config database and remove any reference to the removed namespace from the collections chunks, locks collections:<\/p>\n
use config\r\n\r\ndb.collections.remove( { _id: \"db.Collection\" } )\r\n\r\ndb.chunks.remove( { ns: \"db.Collection\" } )\r\n\r\ndb.locks.remove( { _id: \"db.Collection\" } )<\/pre>\n <\/p>\n
Connect to each mongos and run flushRouterConfig<\/p>\n
After dropping is succeeded, we can change the name of the copy collection:<\/p>\n
db.Collectioncopy.renameCollection(\"Collection\")<\/pre>\n <\/p>\n
As a note :<\/strong> rename collection is not supported in sharded collections.<\/p>\nAfter rename we can create new index on this unsharded collection.<\/p>\n
As you know we need a little down time in this process too, because no new data must be come to existing collection and during drop and rename operations the appliacation can not find the collection. So during these operations we stopped the application.<\/p>\n
It is a challenging operation if your collection size is big so it is always important to decide sharding status and the right shard key before your data grows.<\/p>\n\n
<\/div>\n