HDDS-10564. Make Outputstream writeExecutor daemon threads.

[jojochuang]

What changes were proposed in this pull request?

HDDS-10564. Make Outputstream writeExecutor daemon threads.

Please describe your PR in detail:
HDDS-10383 added new thread pools in the output stream. They should be made daemon threads so that processes don't wait for them when exiting. (The bug is in the command line which does not close FileSystem object. Nonetheless, it would still be a good idea to make these thread pools daemon)

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-10564

How was this patch tested?

Deployed on a real cluster, with HBase on Ozone, repeated the same command to verify that the command no longer hangs.

[xichen01]

It looks like the threadpool in the RpcClient has never been daemon, so is there a similar problem in previous versions?

[adoroszlai]

so that processes don't wait for them when exiting.

Does that mean ongoing write may be dropped and data lost?

[jojochuang]

It looks like the threadpool in the RpcClient has never been daemon, so is there a similar problem in previous versions?

writeExecutor was added by HDDS-10383 and is new.

[jojochuang]

Does that mean ongoing write may be dropped and data lost?

output stream should call close() which will ensure data in the buffer is flushed.
If client doesn't close the output stream then it doesn't guarantee data will be written, I think.

[adoroszlai]

Thanks @jojochuang for the patch, @xichen01 for the review.