Hi all,
As of yesterday I received a mail from Rackspace stating that they are
ending the OSS discount program at the end of Dec. I have spoken to our
Rackspace Account Manager, and there's no offer of some other kind of
discount - we'll need to migrate the workload, or pay for the usage.
Given our heavy use of Rackspace, this is going to take some work to
mitigate.
Many projects are affected by this (e.g. readthedocs[1]) and are
talking about it on Twitter - I'll be doing the same shortly. There's
already a number of companies offering to help (notably the Open Source
Labs [2]), and I'll be contacting them. However we need to decide what
to ask for.
I don't believe we'll find a single sponsor to cover the whole $2k that
we spend at Rackspace each month. Currently on Rackspace we have:
- The core infra - puppetmaster, Jenkins master, and web node
- The permanent Jenkins slaves - 9 of these (by far the largest cost)
- The temporary slaves spun up by BATS and then deleted
- The bandwidth used by the web node - Rackspace claim 6Tb / month, but
I struggle to prove that with my own stats
I think each of these can be handled differently.
Web node / traffic - Right now we have an offer of a sponsored CDN
which I'm trying to finalize, and (separately) and offer of a
mirror+bandwidth from someone else. I'm asking them to run the web node
for us (instead of just a mirror), and with the CDN that should be
acceptable bandwidth.
Core Infra - I will ask Scaleway if we can increase our discount, and
host the Puppetmaster and Jenkins master there. We also have some
hosting with oVirt, that may be possible to increase. Failing that, OSL
or another new sponsor might be able to cover this.
Temporary slaves - if we can find someone to donate some kind of cloud
account (AWS, GCE, Digitalocean, whatever) then this ought to be
fixable in our JJB config - Eric, do I have that right?
Permanent slaves - this is the big one, we simply don't have the
compute power to replace all 9 Rackspace slaves at this time. We are
lucky in that a new slave requires no access to the rest of the infra
(inbound SSH only for the setup), so hopefully we can make a big push
for new sponsors or upgrading the ones we have.
So, I will submit the application to OSL shortly ([3] if anyone else
needs it for another project), tweet out a request for help, and I will
speak to our existing sponsors about more capacity. In the mean time,
if anyone has any contacts or offers to make, let's hear it
Greg
[1] https://twitter.com/ericholscher/status/920396452307668992
[2] https://twitter.com/osuosl/status/920491352634269696
[3] http://osuosl.org/request-hosting/