davehodg at gmail.com
Fri Dec 24 15:42:34 GMT 2010
On 23 Dec 2010, at 14:15, James Laver wrote:
> On Wed, Dec 22, 2010 at 10:41:12PM +0000, Dave Hodgkinson wrote:
>> Either you're doing it wrong or hudson isn't what you want.
>> Rather than flailing around looking for more software fail, why not look
>> at what the problem might be, whether software or procedural.
>> WFM YMMV IANAL.
> You are running hudson for < 20 jobs, I am running it for > 200, a scale
> at which I have to drive it through the groovy console to keep
> everything in sync (which brings in further hate in the form of poor
> documentation and a complete lack of tutorial of the api, not to mention
> having to write groovy code).
Given it's what I'm dong at photobox, dare I say build management is a
full time job?
Also, the "everything in sync" bit worries me. The dependent builds and svn
watching are supposed to do that.
> I think perhaps hudson is not what I want, despite being a big bloated
> java piece of hate, it doesn't actually give me some of the features I
> could use to scale things horizontally appropriately.
Have you looked at parameterised builds? We now build from trunk routinely
but can build .debs of branches or tags when we want too.
> there are certain
> things I like about it, but managing a hudson of this scale is becoming
> too big a timesink when I have other things to be doing.
Played with gridding it? That's my toy in the new year. Getting the f/e
devs and testers working on their seleniums then farming it out onto a
couple of cloud boxes.
More information about the london.pm