On 2013-08-01 21:30, Stefan Milo wrote:
A tool like observium is often used by people in Operations. People in Operations tend to favor stability over new features. The old if it's not broken mantra...
Well then, you probably want to be using Cacti. Judging by the update habits of our existing userbase, they agree with us.
I would hate it if i updated Observium, and because of Toms fat fingers, didn't receive alerts if a critical interface went down.
This is why we have quite strict rules about making sure the poller never breaks.
I'm not saying theres anything wrong with the way you roll out updates. I'm just saying there might be a conflict of interest from you as a developer, and your target audience in Operations.
Well, you see, the fact is that I'm a network engineer, and not a developer. I write it for me to use.
We don't write code and then distribute it for other people to use. I test code as I'm writing it on my own live installations.
There have been 2 kinds of major bugs which have made it in to the svn
a) missing includes b) using new php features too soon
a) has happened a couple of times over the years during code drops when we need to add a lot of new files and one get missed, and for that reason a lot of long time users tend to wait a day or two after a large code drop to update.
b) happened once with a new feature in 5.4
These are always easy to back out of, as you just roll back svn to the previous release and wait for a fix, which is usually done before more than a couple of people have updated.
On a sidenote... Would it be possible to implement a feature that would alert you if you were trying to update a really old install? Something like "Error; you are trying to update from R3193 to R4203. Please update to R4100 before proceeding."
Nope. It's not useful for us, so we won't increase our development workload. If you want to be super cautious, you can maintain a second installation that you test updates on.
Realistically though, it's not really necessary. Most installations I look after update automatically by cron. These installs have historically had far less issues than the installs updated by people who think they know better and follow the stupid "if it's not broken..." mantra.
Yes, I mean you, jduggan.
adam.