R
Rob Fitzpatrick
Guest
In this case it's a dev environment and it's one agent handling one request at a time, so I guess I can handle profiling everything. The only caveat is that the first request after the agent starts is always slow (I assume because it hasn't cached schema, r-code, etc.) and then the subsequent queries are faster. The challenge because of this is that I'd like to profile the *second* request/response, so maybe doing it programmatically is the way to go. I've thought about maybe creating a flag file that the agent can read and using that as a toggle to enable profiling. In other words, start the agent, send a message (which I don't want to profile), consume the response, create the flag file, send another message (which is profiled), stop the agent and flush the data. This programmatic approach is better I think than using -profile in this case, but I'd still like to understand why it isn't working. Also, re "-profiling" in the config file, it seems redundant. I think the fact that I'm specifying "-profile " on the command line should turn the feature on.
Continue reading...
Continue reading...