shard cluster shows data only on 1 node

45 views

Skip to first unread message

Anmol

unread,

May 15, 2018, 5:07:00 AM5/15/18

to mongodb-user

I have defined 2 shards ShardA and ShardB

Shard A has data for ag.table1 from x=1 to 99 and Shard B from x=100 to x=199. however, when i see data distribution, it shows that all of data resides on Shard A Why so??

mongos> db.table1.getShardDistribution()

Shard shard_Repl_SetA at shard_Repl_SetA/xtlinno1vmongo02.xavient.com:34001,xtlinno1vmongo03.xavient.com:34001,xtlinno1vmongo04.xavient.com:34001

data : 8KiB docs : 200 chunks : 3

estimated data per chunk : 2KiB

estimated docs per chunk : 66

Shard shard_Repl_SetB at shard_Repl_SetB/xtlinno1vmongo01.xavient.com:33001,xtlinno1vmongo05.xavient.com:33001

data : 0B docs : 0 chunks : 1

estimated data per chunk : 0B

estimated docs per chunk : 0

Totals

data : 8KiB docs : 200 chunks : 4

Shard shard_Repl_SetA contains 100% data, 100% docs in cluster, avg obj size on shard : 46B

Shard shard_Repl_SetB contains 0% data, 0% docs in cluster, avg obj size on shard : NaNGiB

================

Configuration-->>>

sharding version: {

"_id" : 1,

"minCompatibleVersion" : 5,

"currentVersion" : 6,

"clusterId" : ObjectId("5aeff2b0da17efd745775bf8")

}

shards:

{ "_id" : "shard_Repl_SetA", "host" : "shard_Repl_SetA/xtlinno1vmongo02.xavient.com:34001,xtlinno1vmongo03.xavient.com:34001,xtlinno1vmongo04.xavient.com:34001", "state" : 1, "tags" : [ "ShardA" ] }

{ "_id" : "shard_Repl_SetB", "host" : "shard_Repl_SetB/xtlinno1vmongo01.xavient.com:33001,xtlinno1vmongo05.xavient.com:33001", "state" : 1, "tags" : [ "ShardB" ] }

active mongoses:

"3.6.4" : 2

autosplit:

Currently enabled: yes

balancer:

Currently enabled: yes

Currently running: no

Failed balancer rounds in last 5 attempts: 3

Last reported error: Could not find host matching read preference { mode: "primary" } for set shard_Repl_SetB

Time of Reported error: Tue May 15 2018 11:48:53 GMT+0530 (IST)

Migration Results for the last 24 hours:

6 : Success

{ "_id" : "ag", "primary" : "shard_Repl_SetA", "partitioned" : true }

ag.table1

shard key: { "x" : 1 }

unique: false

balancing: true

chunks:

shard_Repl_SetA 3

shard_Repl_SetB 1

{ "x" : { "$minKey" : 1 } } -->> { "x" : "1" } on : shard_Repl_SetA Timestamp(2, 1)

{ "x" : "1" } -->> { "x" : "100" } on : shard_Repl_SetA Timestamp(1, 2)

{ "x" : "100" } -->> { "x" : "200" } on : shard_Repl_SetB Timestamp(2, 0)

{ "x" : "200" } -->> { "x" : { "$maxKey" : 1 } } on : shard_Repl_SetA Timestamp(1, 5)

tag: ShardA { "x" : "1" } -->> { "x" : "100" }

tag: ShardB { "x" : "100" } -->> { "x" : "200" }

Kevin Adistambha

unread,

May 30, 2018, 9:09:35 PM5/30/18

to mongodb-user

I believe this is the main issue you’re having:

Last reported error: Could not find host matching read preference { mode: “primary” } for set shard_Repl_SetB

It appears that the Shard B replica set does not have a primary. This means that no data can be written to the replica set. In other words, MongoDB tried to balance the cluster, but cannot since Shard B does not accept writes.

One very likely cause if because the Shard B replica set appears to contain only 2 members. This is not a recommended configuration for a replica set, since a replica set depends on its ability to elect a primary member that can accept writes by simple majority. In a 2-member replica set, if one of them is offline, the remaining node will become a secondary to prevent “split-brain” scenario. Thus the minimum of 3 members is the recommended configuration for a replica set. This is described in detail in the Replication page.

To enable the cluster to balance itself, Shard B must be writable (e.g. the replica set having a primary). The best way to solve this is to add another data-bearing member to Shard B replica set.