Summary

https://docs.pixielabs.ai/tutorials/custom-data/dynamic-go-logging/
Pixie Dynamic Logging let users write eBPF uprobes in PXL script, Pixie then deploys the uprobes onto user-specified Pods, and collect the data into Pixie's own data store.

https://docs.pixielabs.ai/tutorials/custom-data/distributed-bpftrace-deployment
Pixie also has Distributed Bpftrace Deployment, which lets users deploy kprobes and uprobes on Kubernetes cluster.

Given the limited capability of the PXL language, Dynamic Logging is used much less than Distributed Bpftrace Deployment. We are looking for feedbacks from the community regarding the future directions of Dynamic Logging.

Suggestions

1. One option is to simply remove Dynamic Logging from Pixie. That greatly reduces our maintenance burden.
2. Another option is to further expand PXL language support for Dynamic Logging to make it more powerful and useful.
3. A hybrid option is to extra the useful part of Dynamic Logging and contributed to BPFtrace, for example, the part of navigating struct variables with DWARF information (already part of BPFTrace https://github.com/iovisor/bpftrace/blob/master/docs/reference_guide.md#4-uprobeuretprobe-dynamic-tracing-user-level-arguments) could get improved with Pixie's current implementation.
4. Any other suggestions?