opencomputeproject
Future roadmap and bugfixes
Engineer:
- session control for user to regain session or close it if browser tab closes out
- compiler node load-balancer architecture
- unit/integration test suite
- go test -cover ready
- _test go modules for all
- security test suite including bits like https://github.com/securego/gosec
- handle and test for unhandled errors via https://github.com/kisielk/errcheck
- eval https://github.com/alufers/mitmproxy2swagger
Port & refactor documentation / issues / discussions from old repo:
- https://github.com/HewlettPackard/osfci/issues
- https://github.com/HewlettPackard/osfci/discussions
- unify https://github.com/opencomputeproject/OSF-OSFCI/search?q=SetConfigType idiom
-- build --
- enhance error handling and checks for deploy script referenced in build_all.sh
- shift to local cache for external deps to avoid breakage
-- contest future work --
Update from discussions with OCP community
Question: will contest comply in future to the OCP spec which is currently under dev?
Answer(s):
contest has support for some stuff already - they're under the exec plugin: https://github.com/linuxboot/contest/tree/main/plugins/teststeps/exec
The plugin was written to run external binaries of any sort, locally or remote thru ssh (see transport), and there's an ocp_output flag for it to look for that output on stdout; otherwise it just says started event , ended event and result is based on exit code (still in the ocp_output=false case) with ocp_output=true, it looks for result in testRunEnd (or whatever the name of that message was), and creates additional events for stuff that happens inside the test (doesn't have measurements iirc) is it correct to assume that eventually contest will also comply
Summary: after OCP publish v1 of spec, maintainers will probably add the rest of the messages parsing
via @gregg-shick
-- GUI , rough thoughts --
- Remove the text window with the tech output – just give a visual to the customer that the key has been written and SUT currently has control
- Move the text window output to external logging for debug if necessary.
-- Acro code needed --
- General observation – key is getting stuck on the SUT
- Control of acro needs to be moved back to controller at end of session, that is not happening now.
- When operation system installer is clicked and an OS selected, we need to 1) move control of acro to the controller if not already 2) write image 3) move acro to SUT
- Need to make sure PV is installed on all controllers as default (we need to script OS setup for each server in the config).
-- Other bugs --
- At the start of the session, power was still present on the SUT from previous session (detectable since BMC WEB popped up and Openbmc was running even though an image for rom/bmc fw was never loaded and connect power was never clicked). Need to ensure power is pulled during session cleanup (@gregg-shick 's opinion).
- We should think about maybe clearing the emulators as part of end session cleanup.
- When selecting OS installer image again we were prompted that we were leaving website. That needs fixing.
- Emulator stability: Don't allow new FW to be loaded in those windows when power is applied to the SUT. Enforce power-off to load a new image for rom or openbmc.
-- security future integrations --
- https://github.com/binarly-io/fwhunt-scan
- https://github.com/e-m-b-a/emba
- https://github.com/fkie-cad/FACT_core
-- security work --
- review and refactor based on https://github.com/securego/gosec results
- deploy and review https://go.dev/blog/vuln