[Board] RE: [Box-Admins] What do I need to ask the SFC for on Wednesday?

[Levente Uzonyi]

On Mon, 24 Sep 2012, Ken Causey wrote:

>> -------- Original Message --------
>> Subject: Re: [Board] RE: [Box-Admins] What do I need to ask the SFC for
>> on Wednesday?
>> From: Chris Cunnington <smalltalktelevision at gmail.com>
>> Date: Mon, September 24, 2012 2:29 pm
>> To: Squeak Hosting Support <box-admins at lists.squeakfoundation.org>
>> 
>> 
>> On 12-09-24 3:21 PM, Ken Causey wrote:
>> >> >So the only thing the Gandi VPS would run for now would be Jenkins. 
>> >> And maybe if we need another server we could get an additional one 
>> >> instead of making this larger? I'm not that experienced with virtual 
>> >> servers.
>> > Then I'm confused why Chris is asking about this.  Isn't Squeak-ci his
>> > baby?
>> >
>> I'm keen to dump my server and get a new one. That means I have to find 
>> a new home for the Jenkins server. I've been clear to the SOB that I 
>> want to make a transition 1 Nov. away from Server Beach. I've said such 
>> things in SOB minutes. So, I'm sort of forcing the issue. The question, 
>> though, is how much do we want to portion to the Gandi server. It's an 
>> open question.
>
> So are you saying the first goal with the server is to run Jenkins and
> the question is what else can we run there?  The answer depends entirely
> on what additional resources are available.  If we can run everything
> else there along with Jenkins, that sounds great to me, but perhaps
> that's not reasonable.  Doesn't Jenkins involve a lot of usage spikes?

During the weekend I installed Jenkins on our VPS at Hetzner. It uses 
about 170 MB memory and 50 MB disk space alone. I added the SqueakTrunk 
job from squeakci.org to it, which updates a Trunk image and runs the 
tests. That job creates about 25 MB of data each build (image file, 
changes file). Since the test fires up a CogVM with a Trunk image, 
additional 20 MB of memory is being used during the build.

So even a single Share Gandi VPS could run Jenkins right now, but we only 
have 3 jobs yet (2 working atm). The goal is to have like 10+ of 
them. I expect every job to be run twice a day on average when automatic 
triggers are installed. So 10 jobs, 2 builds per job per day and 25 MB 
extra disk space per build means, 500 MB new data per day (the VM building 
jobs might require less space per build and we might not keep the 
artifacts of all builds, so this is more like an upper bound).

I think at least 2 Shares (1 CPU, 1GB RAM, 20 Mbps guaranteed 
bandwidth) and 60 GB disk space (10 GB for the OS and environment + 100 
days build history with artifacts) is enough for Jenkins.


Levente

>
> Ken
>
>> 
>> Chris
>