Open-sourcing Teletraan, a code deployment system
engineering.pinterest.com
engineering.pinterest.com
If that's the requirement, a serious question might be: why not just deploy code through your provisioning process?
My personal impression is that the key issue with most systems in this area is that they whitewash, ignore or blur the reality that network topology, security, service codependencies, data security, build and deployment pipelines and so forth need to be managed at some level, may not be the same across all infrastructure, and requirements in these spaces can frequently be nontrivial with competing concerns. Thus, individual solutions typically only function for a subset of cases while ignoring the fact that others exist.
A clear description of the limitations of the approach of the tool would be useful.
Corosync/pacemaker is one mature solution in this area with a focus on HA and the capacity to navigate logical topology changes (due to faults, errors, etc.) in real time to resolve service interdependencies towards a defined 'goal state'. Extremely powerful, with a huge library of 'OCF resource agents' (service definitions) already written.[0] However, it also uses the 'agent on each machine' architecture. Where a small number of services are deployed at large scale in a parallel configuration I believe an agentless configuration methodology with PXE and IPMI power control is a more efficient node management path, though it obviously negatively impacts HA.
PS. 'Features' like rollback and 'hotfix' are arguably negligible if you have a reasonable service build process incorporating versioning.
Because provisioning operates with a much larger timescale, with a different fundamental understanding of the act of manipulating remote state. In a word, it's too slow.
In fact it seems to me that this tool is itself still too slow, like a half-step between Chef/Puppet/Ansible (system provisioning) and Kubernetes (application scaling). The future is clearly in the commodification of hardware and network, i.e. immutable infrastructure and cattle-not-pets.
{{citation-needed}}
PXE boot is a few seconds... restores perfectly known state at all layers of the stack and thus has lots of guarantees.
Slightly off topic, I'm a big fan of continuous deployments, but 500 deployments a day seems excessive. Assuming an 8 hour work day, thats more than 1 deploy every minute. Its great that this system works that seamlessly, but are there really benefits with deploying that frequently?
EDIT: Xeroday helpfully pointed out that there IS a link to code in another blog post in the same series.
So yes, it probably does mean that they are deploying more than once to the same set of servers.