basho / riak-dotnet-client

The Riak client for .NET
Apache License 2.0
76 stars 36 forks source link

enhancement: some kind of Riak logging facility #179

Open guidorice opened 10 years ago

guidorice commented 10 years ago

Suggested to add by OJ :)

peschkaj commented 10 years ago

Sounds good.

Are you looking to capture queries run for MapReduce and 2i? What about Search? Do you just need the query text itself or is something more diagnostic required? Like: "query ABC started at 1; finished at 2;"


sent from a tiny portion of the hive mind... in this case, a phone On Oct 2, 2013 12:50 PM, "mindlube" notifications@github.com wrote:

Suggested to add by OJ :)

— Reply to this email directly or view it on GitHubhttps://github.com/DistributedNonsense/CorrugatedIron/issues/179 .

guidorice commented 10 years ago

Queries and Search would be useful I'm sure, but just at a basic level it would be helpful to see Gets, Puts, bucket, key and r,w parameters!

OJ commented 10 years ago

Thanks for kicking off the issue as requested!

I think we need to make an effort to be clear of what's possible, what's practical and what is it that people need as I don't think they're all the same thing!

There's a lot of functionality in CI, and backfilling logging everywhere is going to be quite an ordeal. I think this is something that should be attempted piecemeal over a long period of time. In the short term, adding logging to the most important calls is probably the go.

I don't think it's practical to add logging at the endpoint because everything that is going out is PBC encoded and won't mean anything to anyone. I also think that trying to get smart and using AOP with a PBC translator is more pain than it's worth.

To start with, let's focus on simple Get(), Put(), Delete(). Map/Reduce can come after that.

The next question, before pushing ahead, is which tool to use. Personally I'm a fan of NLog in the .NET world of logging at the moment, but that would tie CI to one logging tool. The problem with logging is that everyone loves different stuff and moans when they can't their favourites.

So do we go through the pain (and cost in time and perf) of putting an abstraction in place that can do it so that anything can be wired in?

Thoughts?

peschkaj commented 10 years ago

Marking this as deferred so it doesn't delay CI 2.0. Should be able to hit this up further down the line.

lukebakken commented 9 years ago

Related to #218 and #221