Conversation
Add a portable shell script that collects system telemetry for analysis: - Kernel version, architecture, uptime - Board/platform identity (devicetree, DMI) - CPU and memory info - EDAC memory controller data from sysfs - EDAC/MCE error counts from dmesg - rasdaemon SQLite database if available - SGL distro version Outputs a JSON report and tarball suitable for upload to a telemetry platform. Designed to run on any SGL deployment with BusyBox or full coreutils — tested on SGL QEMU ARM64. Signed-off-by: Ramon Roche <mrpollo@gmail.com>
4a255b7 to
98fd8c9
Compare
|
This is a great idea to start the discussion. May I suggest we formally mark it as a draft so we know it is open for discussion? |
|
Thanks @robwoolley I was testing things out in QEMU during the call, but I'm obviously guessing with the EDAC stuff, would be nice to get someone to test this on actual hardware. Maybe a BeagleV fire? ;) |
|
That sounds like a great thing to explore. I see that there is a CoreEDAC IP block that can be loaded onto the FPGA: https://www.microchip.com/en-us/products/fpgas-and-plds/ip-core-tools/coreedac However, I don't see a software interface mentioned in the CoreEDAC user guide or a corresponding EDAC driver for the Linux kernel: https://github.com/linux4microchip/linux/tree/linux-6.12-mchp/drivers/edac I suggest we reach out to Microchip to ask about investigating this further on the BeagleV-Fire and other PolarFire SoC boards. |
Adds a portable shell script that collects system telemetry from SGL deployments, focused on EDAC (Error Detection and Correction) data and hardware error events. The goal is to enable teams running SGL in space or radiation environments to share anonymizable telemetry for aggregate analysis — similar to logs.px4.io but for space-grade Linux.
This is a draft, the upload destination and telemetry platform are TBD. Filed as a starting point for discussion at the SGL SIG.
Sample output: