mongodb pre-splitting sharding测试

来源：互联网发布：炭知天下价目表编辑：程序博客网时间：2024/06/05 19:18

关于mongodb的预分片技术,网上的资料很多，好处是显而易见的，可以减轻因auto-balancer 要split chunk所带来的性能影响，下面介绍Pre-splitting的操作步骤，仅做为测试。

首先，搭建环境，最好是Repl Set+Sharding环境，此次测试是在自己的电脑上做的，两个分片，每个分片3个副本集，要做pre-splitting，最好先计算要分片的collection有多大，平均每条数据有多大，如果不清楚，也可以随便建个coll,插入一条，然后db.XXX.stats()可以看到avgsize，这就是一条数据的大小，单位是byte.

以下是测试过程，本次测试设定chunksize为5M，采用：for(var i=1; i<=200000; i++){
db.test.insert({_id:i,name:"wzw",age:i,uid:i+1});}这条语句插入数据(放到最后操作，此处做为示例)，每条数据大概112bytes，20万条数据大概21M。

本次测试在test库中测试。

use test;

mongos> sh.enableSharding("test");

{ "ok" : 1 }

mongos> sh.shardCollection("test.test",{"_id":1});
{ "collectionsharded" : "test.test", "ok" : 1 }

20万条数据分配在两个分片上，每个分片大概10M（10W条）数据，那么做以下分配：

划分10个chunk，每个chunk 20000条数据；

操作之前先停止自动均衡；

mongos> use test
switched to db test
mongos> sh.stopBalancer();
Waiting for active hosts...
Waiting for the balancer lock...
Waiting again for active hosts after balancer is off...
mongos> use admin
switched to db admin
mongos> for( var y=1; y<10; y++ ) {
... var x=20000
... var prefix = x*y;
... db.runCommand( { split : "test.test" , middle : { "_id": prefix } } );
... }
{ "ok" : 1 }
mongos> sh.status();
--- Sharding Status ---
sharding version: {
"_id" : 1,
"minCompatibleVersion" : 5,
"currentVersion" : 6,
"clusterId" : ObjectId("55c82589acfaad53c0a64845")
}
shards:
{ "_id" : "shard1", "host" : "shard1/127.0.0.1:11111,127.0.0.1:22222,127.0.0.1:33333" }
{ "_id" : "shard2", "host" : "shard2/127.0.0.1:44444,127.0.0.1:55555,127.0.0.1:60000" }
balancer:
Currently enabled: no
Currently running: no
Failed balancer rounds in last 5 attempts: 0
Migration Results for the last 24 hours:
85 : Success
databases:
{ "_id" : "admin", "partitioned" : false, "primary" : "config" }
{ "_id" : "test", "partitioned" : true, "primary" : "shard1" }
test.test
shard key: { "_id" : 1 }
chunks:
shard1 10
{ "_id" : { "$minKey" : 1 } } -->> { "_id" : 20000 } on : shard1 Timestamp(1, 1)
{ "_id" : 20000 } -->> { "_id" : 40000 } on : shard1 Timestamp(1, 3)
{ "_id" : 40000 } -->> { "_id" : 60000 } on : shard1 Timestamp(1, 5)
{ "_id" : 60000 } -->> { "_id" : 80000 } on : shard1 Timestamp(1, 7)
{ "_id" : 80000 } -->> { "_id" : 100000 } on : shard1 Timestamp(1, 9)
{ "_id" : 100000 } -->> { "_id" : 120000 } on : shard1 Timestamp(1, 11)
{ "_id" : 120000 } -->> { "_id" : 140000 } on : shard1 Timestamp(1, 13)
{ "_id" : 140000 } -->> { "_id" : 160000 } on : shard1 Timestamp(1, 15)
{ "_id" : 160000 } -->> { "_id" : 180000 } on : shard1 Timestamp(1, 17)
{ "_id" : 180000 } -->> { "_id" : { "$maxKey" : 1 } } on : shard1 Timestamp(1, 18)
{ "_id" : "user", "partitioned" : true, "primary" : "shard1" }
user.test
shard key: { "_id" : 1 }
chunks:
shard1 1
{ "_id" : { "$minKey" : 1 } } -->> { "_id" : { "$maxKey" : 1 } } on : shard1 Timestamp(1, 0)

可以看到，系统按照预先定义的规则分配了10个chunk，都在shard1上，数据分配很均匀。

mongos> use test
switched to db test
mongos> db.test.stats();
{
"sharded" : true,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0. It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"ns" : "test.test",
"count" : 0,
"numExtents" : 1,
"size" : 0,
"storageSize" : 8192,
"totalIndexSize" : 8176,
"indexSizes" : {
"_id_" : 8176
},
"avgObjSize" : 0,
"nindexes" : 1,
"nchunks" : 10,
"shards" : {
"shard1" : {
"ns" : "test.test",
"count" : 0,
"size" : 0,
"numExtents" : 1,
"storageSize" : 8192,
"lastExtentSize" : 8192,
"paddingFactor" : 1,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0. It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"nindexes" : 1,
"totalIndexSize" : 8176,
"indexSizes" : {
"_id_" : 8176
},
"ok" : 1,
"$gleStats" : {
"lastOpTime" : Timestamp(1439189515, 1),
"electionId" : ObjectId("55c81b008266db61a673c152")
}
}
},
"ok" : 1
}

mongos> sh.startBalancer();

mongos> sh.status();
--- Sharding Status ---
sharding version: {
"_id" : 1,
"minCompatibleVersion" : 5,
"currentVersion" : 6,
"clusterId" : ObjectId("55c82589acfaad53c0a64845")
}
shards:
{ "_id" : "shard1", "host" : "shard1/127.0.0.1:11111,127.0.0.1:22222,127.0.0.1:33333" }
{ "_id" : "shard2", "host" : "shard2/127.0.0.1:44444,127.0.0.1:55555,127.0.0.1:60000" }
balancer:
Currently enabled: yes
Currently running: no
Failed balancer rounds in last 5 attempts: 0
Migration Results for the last 24 hours:
90 : Success
databases:
{ "_id" : "admin", "partitioned" : false, "primary" : "config" }
{ "_id" : "test", "partitioned" : true, "primary" : "shard1" }
test.test
shard key: { "_id" : 1 }
chunks:
shard1 5
shard2 5
{ "_id" : { "$minKey" : 1 } } -->> { "_id" : 20000 } on : shard2 Timestamp(2, 0)
{ "_id" : 20000 } -->> { "_id" : 40000 } on : shard2 Timestamp(3, 0)
{ "_id" : 40000 } -->> { "_id" : 60000 } on : shard2 Timestamp(4, 0)
{ "_id" : 60000 } -->> { "_id" : 80000 } on : shard2 Timestamp(5, 0)
{ "_id" : 80000 } -->> { "_id" : 100000 } on : shard2 Timestamp(6, 0)
{ "_id" : 100000 } -->> { "_id" : 120000 } on : shard1 Timestamp(6, 1)
{ "_id" : 120000 } -->> { "_id" : 140000 } on : shard1 Timestamp(1, 13)
{ "_id" : 140000 } -->> { "_id" : 160000 } on : shard1 Timestamp(1, 15)
{ "_id" : 160000 } -->> { "_id" : 180000 } on : shard1 Timestamp(1, 17)
{ "_id" : 180000 } -->> { "_id" : { "$maxKey" : 1 } } on : shard1 Timestamp(1, 18)
{ "_id" : "user", "partitioned" : true, "primary" : "shard1" }
user.test
shard key: { "_id" : 1 }
chunks:
shard1 1
{ "_id" : { "$minKey" : 1 } } -->> { "_id" : { "$maxKey" : 1 } } on : shard1 Timestamp(1, 0)

可以看到数据被平均分配到两个分片上，现在插入数据：

for(var i=1; i<=200000; i++){
db.test.insert({_id:i,name:"wzw",age:i,uid:i+1});}，通过另一个会话观察：

[root@mongodb ~]# mongostat --port 50000
insert query update delete getmore command flushes mapped vsize res faults qr|qw ar|aw netIn netOut conn set repl time
849 *0 *0 *0 0 850|0 0 263.0M 8.0M 0 0|0 0|0 125k 54k 2 RTR 03:51:22
937 *0 *0 *0 0 938|0 0 263.0M 8.0M 0 0|0 0|0 138k 59k 2 RTR 03:51:23
889 *0 *0 *0 0 890|0 0 263.0M 8.0M 0 0|0 0|0 131k 56k 2 RTR 03:51:24
862 *0 *0 *0 0 863|0 0 263.0M 8.0M 0 0|0 0|0 127k 54k 2 RTR 03:51:25
774 *0 *0 *0 0 776|0 0 263.0M 8.0M 0 0|0 0|0 114k 50k 2 RTR 03:51:26
851 *0 *0 *0 0 852|0 0 263.0M 8.0M 0 0|0 0|0 125k 54k 2 RTR 03:51:27
857 *0 *0 *0 0 858|0 0 263.0M 8.0M 0 0|0 0|0 126k 54k 2 RTR 03:51:28
800 *0 *0 *0 0 801|0 0 263.0M 8.0M 0 0|0 0|0 118k 51k 2 RTR 03:51:29
816 *0 *0 *0 0 817|0 0 263.0M 8.0M 0 0|0 0|0 120k 52k 2 RTR 03:51:30
838 *0 *0 *0 0 840|0 0 263.0M 8.0M 0 0|0 0|0 123k 53k 2 RTR 03:51:31
insert query update delete getmore command flushes mapped vsize res faults qr|qw ar|aw netIn netOut conn set repl time
875 *0 *0 *0 0 876|0 0 263.0M 8.0M 0 0|0 0|0 129k 55k 2 RTR 03:51:32
778 *0 *0 *0 0 779|0 0 263.0M 8.0M 0 0|0 0|0 114k 50k 2 RTR 03:51:33
850 *0 *0 *0 0 851|0 0 263.0M 8.0M 0 0|0 0|0 125k 54k 2 RTR 03:51:34
897 *0 *0 *0 0 898|0 0 263.0M 8.0M 0 0|0 0|0 132k 56k 2 RTR 03:51:35
848 *0 *0 *0 0 850|0 0 263.0M 8.0M 0 0|0 0|0 125k 54k 2 RTR 03:51:36
846 *0 *0 *0 0 847|0 0 263.0M 8.0M 0 0|0 0|0 124k 54k 2 RTR 03:51:37
886 *0 *0 *0 0 887|0 0 263.0M 8.0M 0 0|0 0|0 130k 56k 2 RTR 03:51:38
866 *0 *0 *0 0 867|0 0 263.0M 8.0M 0 0|0 0|0 127k 55k 2 RTR 03:51:39
880 *0 *0 *0 0 881|0 0 263.0M 8.0M 0 0|0 0|0 129k 55k 2 RTR 03:51:40

mongos> for(var i=1; i<=200000; i++){
... db.test.insert({_id:i,name:"wzw",age:i,uid:i+1});}
WriteResult({ "nInserted" : 1 })

mongos> db.test.stats();
{
"sharded" : true,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0. It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"ns" : "test.test",
"count" : 200000,
"numExtents" : 14,
"size" : 22400000,
"storageSize" : 45015040,
"totalIndexSize" : 5625088,
"indexSizes" : {
"_id_" : 5625088
},
"avgObjSize" : 112,
"nindexes" : 1,
"nchunks" : 10,
"shards" : {
"shard1" : {
"ns" : "test.test",
"count" : 100001,
"size" : 11200112,
"avgObjSize" : 112,
"numExtents" : 7,
"storageSize" : 22507520,
"lastExtentSize" : 11325440,
"paddingFactor" : 1,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0. It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"nindexes" : 1,
"totalIndexSize" : 2812544,
"indexSizes" : {
"_id_" : 2812544
},
"ok" : 1,
"$gleStats" : {
"lastOpTime" : Timestamp(1439189515, 1),
"electionId" : ObjectId("55c81b008266db61a673c152")
}
},
"shard2" : {
"ns" : "test.test",
"count" : 99999,
"size" : 11199888,
"avgObjSize" : 112,
"numExtents" : 7,
"storageSize" : 22507520,
"lastExtentSize" : 11325440,
"paddingFactor" : 1,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0. It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"nindexes" : 1,
"totalIndexSize" : 2812544,
"indexSizes" : {
"_id_" : 2812544
},
"ok" : 1,
"$gleStats" : {
"lastOpTime" : Timestamp(1439187391, 1),
"electionId" : ObjectId("55c82518bd50f12a8dd39e0f")
}
}
},
"ok" : 1
}

可以看到，test表被平均分配到两个分片上。

至此，mongodb pre-splitting测试结束，此次测试过程很简单，但是在做预分配chunks时要计算好如何分配，如果没做好，可能导致数据不均衡。当然在生产环境中，可以多做些chunks，为以后数据扩展做准备，并且片键采用hashed方式，不然数据还是不均衡的，本次因测试，所经以只做了10个chunks未做hashed。

0 0