Do you remember the last time you were excited to go somewhere? Were you like a kid saying, “Are we there yet? Are we there yet?”
More likely you were the one getting ‘them’ there. I’m sure it got pretty annoying to keep hearing everyone in the car moaning, wanting some sort of distraction until they got there. Maybe you even had DVD players or some other distraction so you didn’t have to hear the questions.
Now think about how you use Continuous Integration (CI) . Do you have polling of your software configuration management (SCM) tool setup? What about your other process tools, do they poll your SCM as well? Guess what, it’s the same difficult trip. You have tools burdening your working SCM (who slaves away everyday to provide for these other tools!). Every hour/minute/seconds some tool is asking your SCM tool “Are we there yet?”. Doesn’t make a lot of sense does it? Think about how many tools you have running daily. You might have multiple CI machines setup, multiple reporting machines, deployment machines all asking the same question over and over.
Quite a burden.
Now think about what your users are doing during this time. They are looking for the same distractions you might have given the kids. They are off emailing their buddies about the latest game, or watching AccuRev on Youtube (OK, maybe something even more enjoyable on YouTube). Doesn’t sound very productive, yet these automation tools were meant to do just that, increase productivity.
So what do you do? Well, a lot of tools allow you to flip the model. Push your information, don’t pull it.
If you have a large enough development group this won’t be enough. Maybe there really are code changes going into the system every minute. If you also increase the granularity of the information going to your CI and reporting tools, these tools can then decide the correct time to be a burden.
You can also reduce the frequency that you give out the same information. If you have several tools (or stages in a tool) depending on the same answer, they can get the answer from a secondary source that gets populated once.
You also want to be diligent about your monitoring. If you see a periodic load on your tools, justify the load. If it looks strange that Kevin is checking the history of a stream every minute, it probably is. If instead you saw CIMonitor as the user it would be more explicit. And it would be obvious that this should change.
And really, changes like these apply to any tool you are using. Do you really need updated reports everytime a bug is fixed? What about every day instead? If you reduce the burden on your tools to a ‘necessary’ level, then they can be further used to answer other questions.
Are we there yet?
