Managing Director of cloud consultancy EngineerBetter. Lapsed jiu-jitsu instructor, moustache-wearer.

London, England
Joined February 2013
Daniel Jones retweeted
Join Lade and Chisa of @EngineerBetter at Agile Cambridge to hear about their experiences of making experiments inside the organisation work: bit.ly/3wbQZZN #agilecam
1
2
Daniel Jones retweeted
I wasn't involved in the original service-catalog k8s implementation of OSBAPI can't comment on that. Crossplane.io and Tanzu Application Platform both use the Claim analogy for Services using CRDs (non-OSBAPI though).
1
1
3
Replying to @DanielJonesEB
I'd argue many organizations should start this way, definitely not just regulated industries.
1
1
Replying to @DanielJonesEB
We have discussed it internally a lot. Developers don't need cf push, they need git push, that will run the tests for them, promotion and everything related. cf push would only be a single line in the pipeline.
2
1
4
Replying to @DanielJonesEB
I worked with a customer when I was at Pivotal to do this. Opinionated Concourse pipelines published by a small team. Lunch and learns, office hours and a stack overflow like confluence page.
1
2
I remember the Diego feature flag magic :-) meanwhile I futz with my pipeline to add NSX segments and NAT rules to my Kubernetes cluster and I recall “PKS and BOSH did this all for me”.
4
6
Starting the year on a good note with a beautiful explanation of @concourseci by @deniseyu21 🌻
1
3
1
22
Replying to @DanielJonesEB
FWIW I did that (automated regression testing and version control) 27 years ago already. 😇
1
1
Seeing this happen was a "pivotal" moment in my career (pun intended)
1
3
You either die a hero or live long enough to see yourself become the villain
1
8
Daniel Jones retweeted
Replying to @DanielJonesEB
Finally, another key issue related to lifecycle is the fact that containers lifecycle is at least one order of magnitude smaller than VMs. This means changes to BOSH director would also be related to timing and allowing perhaps plugins to delegate some decisions.
1
1
Daniel Jones retweeted
Replying to @DanielJonesEB
Lastly, the magic of our (DK and I) attempt was using Kube’s Pod Disruption Budget (PDB) and that explains its success and failure. Success due to using a Kube primitive to deal with pod lifecycle and failure because BOSH is designed to own that lifecycle
1
1
Daniel Jones retweeted
Replying to @DanielJonesEB
Additionally Sandy Cash (IBM) and team attempted another (separate) effort and presented their results at CF Summit in Basel, Switzerland. If you search YouTube you should find video
1
2
Daniel Jones retweeted
Replying to @DanielJonesEB
Well it’s been a while. What I remember is that our first effort ended up with issues dealing with how the BOSH director wants to be in full control of the VMs (pods) and can be in conflict with Kube. Of course changes to the director could have alleviated… but uphill an effort – at San Jose, CA
1
1
This might relate to one of the disgusting secrets of our industry, which is that product success only depends very weakly on software development effectiveness.
2
1
1
5
Daniel Jones retweeted
Replying to @DanielJonesEB
Check out the Tanzu Application Platform sessions this week at VMworld, I think you’ll like where this is going.
3
10
Daniel Jones retweeted
Replying to @DanielJonesEB
More like the successor to cloud controller ng ;) brings the pieces together to make a modular , extensible cf push
1
1
3
Daniel Jones retweeted
Replying to @DanielJonesEB
I would invert this. Platform concerns like logging, monitoring, credential mgmt, etc also need to be solved for workloads outside CF. Ppl need CF to integrate well with best in class solutions for those concerns rather than bring its own separate solution, no matter how good.
2
4
Daniel Jones retweeted
Replying to @DanielJonesEB
As OSS project it may not effectively. As stewarded by Pivotal this was my full time job evidenced by fact that 100% of PCF deals had 3rd party ISV attached. Upstream could have benefitted from buildpack and service broker marketplace like Docker hub model.
1
1
4