Device Development¶
Process Definiton¶
The contributor(s) requests with CTRL to participate in this contribution scheme
- A member of CTRL (most often ICI) discusses the request with the contributor in detail. Importantly, the following points should be clarified in a round where expertise specific to the project exists
scope of contribution, e.g. what devices, what hardware, special requirements (connectivity)
level of support from CTRL anticipated during development: pair-programming, recurrent review, final review
caveats, rules of conduct, failure scenarios that could impact unrelated services
support contacts, under which the DOC can reach the contributor are defined
if there are considerable reservations by CTRL if the concept is feasible, these need to be addressed before proceeding with further steps.
TODO: add special cases which are especially critical, and need special consideration when being touched, e.g. pulse patterns, MPS, high data rate.
- The CTRL contact identifies a suitable development host for the project and sets up infrastructure
minimally, the device server name needs to indicate that this is a non fully DATA supported system
these hosts are normally excluded from CTRL driven deployment, but it is expected that contributors enable mandatory updates, e.g. by assuring their code is versioned.
Question: do we want a fixed set?
- Access to the development host is provided to the contributor(s). General access policies apply, i.e. if framework modification are intended, no support is given to the service, if only devices are developed the framework installation is supported by DATA
the CTRL contact trains the contributor in the gitlab system, code review and deployment tools, as well as how to hot-develop on the host
- Contributors develop their project
while under development, any devices running on the device server identified in (3) are not supported by DATA, advice and help is on a best effort basis
developers can ask for recurrent review, or pair development as agreed on in (2)
TODO: how to we assure that all operators are informed that the device in the reduced-support state.
When the project stabilises, the contributor can choose to transfer ownership to CTRL
Add a list of requirements? E.g. setup.py needs to work with deployment. PEP8. The devices has been successfully operated in the intended scenario.
A timeline for the transition is agreed upon by CTRL and the author, after CTRL has gotten an initial overview of the device state
a mandatory review by CTRL is made, and the contributor is expected to participate in the review process in a constructive fashion, it is the contributor who is responsible for following up on reviewer requests. Only after a successful review can the device be supported by CTRL. Ideally, another member of the contributor group joins this review.
the device is tagged
the tagged devices is migrated to a host / device server that does not have any support restrictions, and importantly properly deployed - no “dirty” tag!
deployment, bug fixes and new tags are now created by CTRL. The contributor is encouraged to contribute, but they cannot hot-develop anymore
at any time, should the contributor wish to hot-develop again, ownership is transferred back to them, support is restricted, and the device is moved to a device server / host which indicates reduced support. This restarts the cycle at (1) with the base for development being the last support version. This might also be necessary for a short time.
Important notes¶
a contributed device running on a non-reduced support server is not supported at all.
a device is either fully supported or in reduced support, not in both states
if CTRL needs to take ownership of a device, because a device is e.g. needed by another instrument, it is expected that the original author facilitates the process
if CTRL needs to take ownership of a device, because the author
Devices which have a detrimental impact on services unrelated to itself, may be shutdown by CTRLs by the means they deem necessary, to preserve overall system integrity
TODO: responsibility for restart
TODO: framework deployments.