# Run Sysreport ## In this learning path - [Introduction](https://learn.arm.com/learning-paths/servers-and-cloud-computing/sysreport/) - [Before you begin](https://learn.arm.com/learning-paths/servers-and-cloud-computing/sysreport/before-you-begin/) - [Run Sysreport](https://learn.arm.com/learning-paths/servers-and-cloud-computing/sysreport/run-sysreport/) - [Analyze the results](https://learn.arm.com/learning-paths/servers-and-cloud-computing/sysreport/analyse-results/) - [Next Steps](https://learn.arm.com/learning-paths/servers-and-cloud-computing/sysreport/_next-steps/) ## About Sysreport Sysreport is a system capability reporting tool that gives application developers a quick summary of the performance features available on a Linux system. It reports the system configuration in a way that is focused on performance analysis. The tool is aimed at anyone trying to profile performance on Arm Linux systems; this includes cloud instances, bare metal servers, small developer boards, and Raspberry Pi devices. After running Sysreport, a summary is generated and displayed on screen. Information reported includes hardware configuration, operating system configuration (build time), and other configuration settings that can be changed. By default, the tool displays advice about possible configuration changes you can make in order to improve the ability to collect performance information. You can disable the advice using the `--no-advice` option. Example use cases for Sysreport include: - As a developer, you want to know if a cloud instance supports a particular performance feature so that you can debug a performance problem - As a developer, you want a quick, single-page summary of a system’s performance configuration so that you don’t need to run multiple commands manually to gather what you need - As a developer, you would like to know suggested configuration changes you can make to a system so that you can improve your ability to collect performance information ## Run Sysreport Run Sysreport with the following command: ```bash python3 sysreport.py ``` Below is the output from an Arm Neoverse N1 System Development Platform (N1SDP), a reference board based on the Neoverse N1 core, running Ubuntu 20.04.6 LTS: ``` __output__ System feature report: __output__ Collected: 2024-03-20 12:24:18.102663 __output__ Script version: 2024-03-20 11:42:11.356472 __output__ Running as root: False __output__ System hardware: __output__ Architecture: ARMv8.2 __output__ CPUs: 4 __output__ CPU types: 4 x ARM Neoverse N1 r1p0 __output__ cache info: size, associativity, sharing __output__ cache line size: 64 __output__ Caches: __output__ 4 x L1D 64K 4-way 64b-line __output__ 4 x L1I 64K 4-way 64b-line __output__ 4 x L2U 1M 8-way 64b-line __output__ 2 x L3U 1M 8-way 64b-line __output__ 1 x L4U 8M 16-way 64b-line __output__ System memory: 15.6G __output__ Atomic operations: True __output__ interconnect: CMN-600 x 1 __output__ NUMA nodes: 1 __output__ Sockets: 1 __output__ OS configuration: __output__ Kernel: 5.18.4+ __output__ config: /boot/config-5.18.4+ __output__ 32-bit support: True __output__ build dir: /lib/modules/5.18.4+/build __output__ uses atomics: True __output__ huge pages: False __output__ transparent: madvise __output__ MPAM configured: False __output__ resctrl: False __output__ Distribution: Ubuntu 20.04.6 LTS __output__ libc version: glibc 2.29 __output__ boot info: ACPI __output__ KPTI enabled: False __output__ Lockdown: landlock, lockdown, yama, integrity, apparmor __output__ Mitigations: spectre_v2:CSV2, BHB; spectre_v1:__user pointer sanitization __output__ Performance features: __output__ perf tools: True __output__ perf installed at: /usr/lib/linux-tools/5.18.4+/perf __output__ not a dynamic executable __output__ perf with OpenCSD: False __output__ perf counters: 6 __output__ perf sampling: SPE __output__ perf HW trace: ETM __output__ perf paranoid: 0 __output__ perf in userspace: disabled __output__ interconnect perf: True __output__ /proc/kcore: True __output__ /dev/mem: True __output__ not a dynamic executable __output__ Actions that can be taken to improve performance tools experience: __output__ perf tools cannot decode hardware trace __output__ build with CORESIGHT=1 ``` Here is the output from an AWS EC2 cloud instance, running Ubuntu 22.04.3 LTS: ``` __output__ System feature report: __output__ Collected: 2024-03-21 13:37:51.248565 __output__ Script version: 2024-03-21 11:50:23.254919 __output__ Running as root: False __output__ System hardware: __output__ Architecture: ARMv8.4 __output__ CPUs: 16 __output__ CPU types: 16 x ARM Neoverse V1 r1p1 __output__ cache info: size, associativity, sharing __output__ cache line size: 64 __output__ Caches: __output__ 16 x L1D 64K 4-way 64b-line __output__ 16 x L1I 64K 4-way 64b-line __output__ 16 x L2U 1M 8-way 64b-line __output__ 1 x L3U 32M 16-way 64b-line __output__ System memory: 30.8G __output__ Atomic operations: True __output__ interconnect: x 1 __output__ NUMA nodes: 1 __output__ Sockets: 1 __output__ OS configuration: __output__ Kernel: 6.5.0 __output__ config: /boot/config-6.5.0-1015-aws __output__ 32-bit support: True __output__ build dir: /lib/modules/6.5.0-1015-aws/build __output__ uses atomics: True __output__ huge pages: False __output__ transparent: madvise __output__ MPAM configured: False __output__ resctrl: False __output__ Distribution: Ubuntu 22.04.3 LTS __output__ libc version: glibc 2.35 __output__ boot info: ACPI __output__ KPTI enabled: False __output__ Lockdown: landlock, lockdown, yama, integrity, apparmor __output__ Mitigations: spectre_v2:CSV2, BHB; spec_store_bypass:Speculative Store Bypass disabled via prctl; spectre_v1:__user pointer sanitization __output__ Performance features: __output__ perf tools: False __output__ perf installed at: /usr/lib/linux-tools/6.5.0-1015-aws/perf (does not exist) __output__ ldd: /usr/lib/linux-tools/6.5.0-1015-aws/perf: No such file or directory __output__ perf with OpenCSD: False __output__ perf counters: None __output__ perf sampling: None __output__ perf HW trace: None __output__ perf paranoid: 4 __output__ perf in userspace: disabled __output__ interconnect perf: None __output__ /proc/kcore: True __output__ /dev/mem: True __output__ Actions that can be taken to improve performance tools experience: __output__ perf tools not installed __output__ install perf package __output__ or build from kernel sources __output__ System-level events can only be monitored by privileged users __output__ set kernel.perf_event_paranoid=0 __output__ Hardware perf events are not available __output__ ensure APIC table describes PMU interrupt __output__ non-invasive sampling (SPE) not enabled __output__ ensure APIC table describes SPE interrupt __output__ kernel module arm_spe_pmu.ko must be built __output__ hardware trace not enabled __output__ rebuild kernel with CONFIG_CORESIGHT __output__ ensure ACPI describes CoreSight trace fabric ```