events log

Is this something someone would be willing to take on as a paid request? We'd like to be able to graph events which might include things like slowdowns in the system, 3rd party supplier outages, specific machine outages or upgrades. We need start/stop times on each and we'd like this to be a graph that can be displayed along with other graphs - so you can visually correlate events with data abnormalities. We can supply examples of what we are seeking.

On Jan 11, 2013, at 8:01 AM, Carolyn Bray <email address hidden> wrote:

> We'd like to be able to graph events which might include things like
> slowdowns in the system, 3rd party supplier outages, specific machine
> outages or upgrades. We need start/stop times on each and we'd like
> this to be a graph that can be displayed along with other graphs - so
> you can visually correlate events with data abnormalities. We can
> supply examples of what we are seeking.

Something like this? http://codeascraft.etsy.com/2010/12/08/track-every-release/

Ah, actually - I think I understand. At Etsy, we use "Display as Infinite" to display deploys on our graphs. I wouldn't recommend using it for everything you're listing, though, as those lines can get very confusing quickly.

However, we're working on a larger metric correlations engine that should be able to robustly do what you want. It is not ready for release yet, but it will be open sourced within the coming months.

Esty's solution is good if I only have one event that I want to monitor. We have multiple servers, software updates on different servers at different times, outages on different systems or even 3rd party outages that we need to identify separately. Ours is a complex system with many different functions/features. From the Etsy solution, it appears that 1) we'd have lots of event databases to try to differentiate the data events 2) we could not have an event that isn't momentary - so trying to track an outage that last hours would be impossible.
The current solution we have is a basic timeline that can handle multiple events and do points (displayed as large dots) for the instantaeous and lines linking the dots for a particular "event". So, each event is given a start/stop time. those times are identical for instananeous events and different for long events.

With https://github.com/Dieterbe/anthracite and https://github.com/Dieterbe/graphitejs (using the flot backend, not the png backend) you can do it out of the box. it's not hard to add support for end times and show region markings instead of points-in-time (the feature exists in flot, but haven't had the time to bind it in graphitejs). you can easily use colorcoding and text annotations to make sense out of the events.

the "display as infinite" stuff is very limited (only with png output, so no client side interactivity, no text annotations)