Jim G's dashboard

A comment on the question "Possible to have one paid host and one free host?" in New Relic:

Hi Stephen,
For single host sites you'd only need one account for each site. It's only for folks with multiple hosts where some are paid vs. Lite that they need two accounts. You can have as many hosts as you need in an account, but the account can only be at one subscription level. Please let me know if I haven't correctly understand the problem you have.

Thanks – Jim G, on January 23, 2009 20:52

Jim G replied on January 22, 2009 14:12 to the question "3rd app not showing up or logging" in New Relic:

Hi tinomen,
This error occurs when a thread does not have a clean entry/exit from a method that we instrument. Normally, there is nothing your application itself can do to make this happen, so I wonder if there is a funny interaction with the COW mongrel gem. Perhaps forked mongrels put our instrumentation into a bad state.
We can probably add a little debugging to help narrow down the problem if you wish. What version of the new relic agent are you running? I can send you some patches to try out.

- Jim

Jim G replied on January 15, 2009 14:29 to the question "RPM not working with Rails 2.3 edge" in New Relic:

Hi laktek,
For some reason it appears we are not detecting your http frontend. What are you using (ie mongrel/passenger/thin/webbrick/etc)?

- Jim

Jim G replied on January 03, 2009 15:53 to the question "How can we suppress sending our application source code to New Relic's servers?" in New Relic:

Hi Matt,
Now I understand the cause for your concern. The good news, as you guessed, is that we don't actually send any source to our website when you're running the agent in "developer mode". All we do is render source files that are stored on your local computer.

In fact, when running in developer mode, we send no data whatsoever to the RPM website. Everything you see is stored in memory on your local computer.

Let me explain a bit more. Our agent has a single data capture mechanism. In order to gain visibility into what your app is doing, we instrument controller actions, view rendering, activerecord, and a few others. This capture mechanism creates aggregate metrics and transaction traces. Now, here's the difference between developer mode and production mode. In developer mode, the aggregate metrics are dropped and the last 100 transaction traces are available for viewing on your local machine (ie http://<localapp>/newrelic). In production mode, the aggregate metrics are sent to the RPM website and only the slowest transaction trace is sent to RPM (assuming you have a silver or gold subscription). Transaction traces in production mode are a subset of those seen in developer mode in the following ways: a) not every node has a stack trace, b) not every SQL is EXPLAINed, c) no source is available.

Hope that helps!

- Jim</localapp>

Jim G set one of Jim G's replies as an official response to "How can we suppress sending our application source code to New Relic's servers?" in New Relic

Jim G replied on January 02, 2009 22:24 to the question "How can we suppress sending our application source code to New Relic's servers?" in New Relic:

Hi Matt,
You can suppress source code from being sent by setting the "capture_source" option of "error_collector" to false. More about this can be found in our Agent configuration doc posted at:

http://sites.google.com/a/newrelic.co...

Just to be clear, we only capture a few lines of source for ERB errors - the same thing you see on the Rails error page. Let us know if you have any more questions. Thanks!

- Jim

Jim G replied on December 30, 2008 02:32 to the question "Diagnosing high CPU use" in New Relic:

Hi rmatei,
Your comment about the time consumption page is correct. It finds "wall clock" burners, not necessarily CPU burners. If you have RPM silver or gold, you can use transaction traces to find CPU burners. Each transaction trace shows how much CPU was burned for that transaction. Look for transaction traces for common actions and see which consistently burn high amounts of CPU.
For what it's worth, this use case is one we're investigating a better solution for.

- Jim

Jim G replied on December 23, 2008 14:57 to the question "3rd app not showing up or logging" in New Relic:

I think we need to see more debug log output. What would be helpful is to:

1) Delete log/newrelic.*.log
2) Restart your app
3) Let it run for 10 minutes - make sure you exercise the app so that all 5 instances get traffic

Then send us the full log/newrelic.*.log files again. At debug logging level we may be able to figure out why some of the mongrels are not connecting.

- Jim

Jim G replied on December 20, 2008 03:21 to the question "3rd app not showing up or logging" in New Relic:

Hi tinomen,
Well, our agent does work with Passenger, which is similar in its forking behavior, so there is at least some chance the agent will work if you can get it to identify the environment. You probably already know this, but the file local_environment.rb is where we check for known environments. Good luck!

- Jim

Jim G replied on December 17, 2008 00:59 to the problem "2.7.4 causes Error calling Dispatcher.dispatch" in New Relic:

Hi Jonathan,
I'm sorry you're having this problem. I'll have our agent engineer look into it in the morning.

- Jim

Jim G set one of Jim G's replies as an official response to "Activity missing from production" in New Relic

Jim G replied on December 16, 2008 03:49 to the problem "Developer mode losing samples" in New Relic:

Hi rmatei,
It sounds like you might be running Passenger - in development mode all data is kept in process memory and if you have more than one process handling requests you'll see the error you mention.

There is indeed a way to see SQL traces in production. It's a very cool feature of RPM called Transaction Tracer. If you'd like, we can set you up with a free week to try it out. It's available at our Silver and Gold subscription levels. Send an email to support@newrelic.com and we'll get you setup.

- Jim

Jim G replied on December 16, 2008 03:04 to the problem "relic adds significant load to our webservers" in New Relic:

Hi Michael,
Let me know if there is anything else we can do for you. If you're seeing <= 10% overhead from our Agent, then I would say that is expected.

- Jim

Jim G replied on December 16, 2008 02:52 to the problem "Activity missing from production" in New Relic:

Hi mtron,
Our data collection tier was under heavy load at 7:34am to 8:12am due to a batch job that normally runs at midnight but was restarted due to a failure. When our collection tier is busy, data will aggregate in your application instances and eventually will report when the collection tier is available again. This aggregation does not affect the speed or responsiveness of your application, so while some performance data was lost, your application would not have been affected.

- Jim

Jim G replied on December 15, 2008 16:18 to the problem "Mongrel instances disappearing from configured application" in New Relic:

Hi Olly,
This should be easy to troubleshoot because we have pretty decent logging in the Agent. The best thing to do is look at RPM and find a Mongrel you think is reporting but isn't showing up in RPM. Grab the newrelic.log file for that mongrel (each mongrel has a separate log file). Take a look at the bottom of the log file to see if there are any errors. If so, then email us the log file (support@newrelic.com). If there are no errors, then change our Agent's log level to "debug" from "info" (in config/newrelic.yml). At debug logging, all communication with RPM is logged and we'll be able to find out what's wrong for sure.

Thanks,
Jim

Jim G replied on December 15, 2008 16:12 to the question "Can you provide the "referring URL" for error traces?" in New Relic:

Hi John,
Yes, this would be easy. I'll put in on our backlog. Thanks!

- Jim

Jim G started following the idea "Grouping stories together" in Pivotal Labs.

A comment on the problem "Getting MySQL error since installing new relic" in New Relic:

No problem! I'm relieved the problem wasn't ours. – Jim G, on December 10, 2008 16:19

A comment on the problem "Threshold values lower than 1 (such as 0.5) are not currently allowed" in New Relic:

Good point BJClark. We'll take a look at this and see what kind of solution we can come up with. – Jim G, on December 08, 2008 16:23

A comment on the idea "Email Alerts for Error Thresholds!" in New Relic:

Hi SportSpyder - with Silver you can define a threshold for errors on the dashboard that when crossed, will trigger an incident which will show up through RSS. The content of the errors won't be in the feed message, but at least you'll know to look at RPM. Is this good enough or did you have something else in mind?

- Jim – Jim G, on December 05, 2008 02:12