Update the limitation of multiple servers binding to the same http/grp…

[krishung5]

…c port

[dyastremsky]

Are we sure this is the best place for this documentation? I think users usually check our documentation (e.g. .MD files) rather than the command line arguments help screen. Especially since it's quite long. If we want users to know about and understand these options, it might make sense to also include them in documentation.

[krishung5]

@dyastremsky I might be wrong but I couldn't find any documentation regarding anything about http/grpc port except in main.cc. It seems like we ask users to use tritonserver --help to see those options that are not in the MD files. That's why I only updated main.cc, but I'm willing to add those options to the document if this makes more sense.

[Tabrizian]

"If you plan to use this option as a way to load-balance between different triton servers, the same model repository or set of models must be used for every server."

Note that this feature only supports stateless models.

I think it might be better to remove this sentence since the customers may figure out a way to control which server gets the requests and address this limitation.

[krishung5]

I see. Updated the document, thanks for the comment!

[dyastremsky]

@dyastremsky I might be wrong but I couldn't find any documentation regarding anything about http/grpc port except in main.cc. It seems like we ask users to use tritonserver --help to see those options that are not in the MD files. That's why I only updated main.cc, but I'm willing to add those options to the document if this makes more sense.

I see, okay. Sounds like we're being consistent here then.

[dyastremsky]

Here and below:

• Triton should be capitalized.
• Load balance should not have a dash.

[krishung5]

Thanks for the catch! Updated.