Prevent application crash on init time

[aminmaghsodi]

Prevent application crash on init time when less than all graphd servers are available.
In the current version of nebula-java, if any of the graphd servers mentioned in config, is not accessible, the app going to crash.

[Aiee]

I believe it is reasonable to explicitly tell the user if any graphd server is down in the initialization phase, so the user will be notified at an early stage about the problem. What do you think?

Update:
I re-considered this change and found it could be helpful for a use case like this:
For a long-running cluster consisting of 5 graphd nodes, there are 2 graphd nodes that went down but the cluster is still functional, a user may try to connect the clustering filling all 5 graph addresses when init the client.

[Aiee]

@aminmaghsodi Please check https://github.com/vesoft-inc/nebula-java/runs/5492274297?check_suite_focus=true to fix the lint.

[aminmaghsodi]

@aminmaghsodi Please check https://github.com/vesoft-inc/nebula-java/runs/5492274297?check_suite_focus=true to fix the lint.

Done, Thanks @Aiee

[aminmaghsodi]

I believe it is reasonable to explicitly tell the user if any graphd server is down in the initialization phase, so the user will be notified at an early stage about the problem. What do you think?

Update: I re-considered this change and found it could be helpful for a use case like this: For a long-running cluster consisting of 5 graphd nodes, there are 2 graphd nodes that went down but the cluster is still functional, a user may try to connect the clustering filling all 5 graph addresses when init the client.

Thanks for your reply and re-consider your belief. If we say nebula graph is designed to be (semi) highly available and scalable, it means there shouldn't be any minor problem affects client, when other components can cover that problem. for example in this problem, when a graphd server is down, another graphd server is present and up in the cluster to cover requests. Consider a situation, when one of graphd services is down at the start time. what is happenning? Actually nothing. because n-1 graphd servers are running. this feature is considered in storaged servers very good. also it is considered in calling graphd service at the runtime. at the start time if some storage servers are down, client does not affect.
If you see, all other scalable data stores (like hadoop, hbase, cassandra, redis and so on) are designed in way that client, should not be affect in minor problem situations like "One of the REST apis of a huge cluster is down!"

[aminmaghsodi]

I need an approve. anybody to help? :D

[Aiee]

LGTM. Good job!