A consistency of split shard bug in v7.3.1-release

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

A consistency of split shard bug in v7.3.1-release

zhenyuan wei
Hi all,
    I have a collection1  with 8 shards,each shard‘s  replicationFactor=1.
I have an application adding 6000w document with infinite retry if any
Exception catch.
That is to say, finally it should be found 6000w docs when query=*:*.
    Normally, all things good, but if in the same time, a SPLIT SHARD
request happen concurrently,
it would get numFound < 6000w docs finally。
    I found that if  sub shard in "active" status,it would throws exception
“*Request says it is coming from parent shard leader but we are in active
state*” , but parent shard do not throws this exception to client,so client
do treat this request as SUCCESS。
    I write a UT and can reproduce this Exception. Any one know why?