Harris.Chu

just a test for batch update

why the graph is changed after a while

1

https://user-images.githubusercontent.com/1726587/178256240-e9c8f1e8-d0cf-4062-b7b9-469dbe0839db.mov 8

当缩容比如 6 -> 3 需要先迁移走要 remove 节点的 leader。 否则有可能在 balance 过程中，leader 不断切换。 PS: 未来也可以考虑在内核版本里做类似功能。

error for ccore

1

it's a little confused to use `nerrors.CodeError` if the error code is `ErrorCode_E_PARTIAL_RESULT`, and the data is lost because the error code is not ErrorCode_SUCCEEDED. 

support to get result response byte size

2

fix https://github.com/vesoft-inc/nebula-go/issues/293 please note: when invoke `GetByteSize()`, it would serialize the thrift struct. if the struct is a huge struct, it may cost much more time and memory.

Need an interface to get the byte size of result set

2

sometimes user want to get the result set byte size, i.e. how many bytes are returned by graphd. as it's not elegant if we modify thrift code, just add an...

enable to set thrift http response buf limit

1

depends on : https://github.com/vesoft-inc/fbthrift/pull/4 default response buffer limit is 16MB, if the response buffer is larger than that, it would renew a buffer. could let use to change the limit...

1. need to test performance after we replace the protocol. 2. it could prevent server crash if invalid data length

ulimit -n is too low in arm docker

2

``` docker run -it --rm centos:7 /bin/bash -c 'ulimit -n' # arm 1024 # x86 1048576 ``` we should change the ulimit conf in docker file