I am currently expanding a group of shard
Previously, it was 3 shard, 7 billion data, 1.2T, 2.4.3
When I expanded it to 6 shard, I found several problems:
-
Looking at mongos through mongostat, we found that the number of writes per second is about 500, while the number of writes allocated to the newly added shard is about 1-5, that is, the number of writes on the old shard is still very large, while the number of writes on the new shard is very low, so chunk of newly written data cannot be evenly distributed.
-
Move chunk often has the problem of too big. I checked that chunk size is 32M. I adjusted it to 64M and ran all night. just now I looked at the balancer and this error still occurred. what is even more surprising is that too big was reported in the error log.
estimatedChunkSize
In fact, it is only 20M~50M, smaller than my chunk size. -
Move chunk is very, very slow, usually about 20 in one night. Every piece of my data is very small, is:
{"_id":BinData(x,xxxxxxxxx),stat:1}
Such things.
Please answer, thank you