[SOLVED] Proxmox (Debian) server CPU spikes

[malabida]

I have been having a problem with my proxmox server for several days. The installation consists of a clean Debian on which I have installed proxmox using their instructions. The problem is that every once in a while the CPU goes up to 20-60% for no apparent reason. In htop no process is consuming that amount of CPU and in top the sys goes up but not the total consumption.

I have done different tests with PSN and perf. The first tool didn't give me much information but the second one showed me different symbols that are causing a high consumption (I guess).

I disabled swap and set tuned to `throughput-performance` and nothing has changed.

I attach the image and the text response from the `perf report` after the record command.

Samples: 326K of event 'cycles', Event count (approx.): 63571299029
  Children      Self  Command          Shared Object                            Symbol
+   14.12%     0.00%  swapper          [kernel.kallsyms]                        [k] secondary_startup_64_no_ver
+   14.08%     0.05%  swapper          [kernel.kallsyms]                        [k] cpu_startup_entry
+   13.95%     0.23%  swapper          [kernel.kallsyms]                        [k] do_idle
+   13.53%     0.00%  swapper          [kernel.kallsyms]                        [k] start_secondary
+    9.19%     0.02%  swapper          [kernel.kallsyms]                        [k] cpuidle_enter
+    9.09%     0.14%  swapper          [kernel.kallsyms]                        [k] cpuidle_enter_state
+    4.66%     1.31%  swapper          [kernel.kallsyms]                        [k] acpi_idle_do_entry
+    4.39%     0.09%  swapper          [kernel.kallsyms]                        [k] acpi_idle_enter
+    3.58%     0.00%  CPU 0/KVM        qemu-system-x86_64 (deleted)             [.] 0x000055a52cfd9700
+    3.58%     0.00%  CPU 0/KVM        libc-2.31.so                             [.] ioctl
+    3.58%     0.00%  CPU 0/KVM        [unknown]                                [k] 0x7570632d34365f36
+    3.58%     0.00%  CPU 0/KVM        [unknown]                                [k] 0x0000000000000190
+    3.58%     0.00%  CPU 0/KVM        [unknown]                                [k] 0x000055a52fb15d10
+    3.50%     0.00%  CPU 0/KVM        [kernel.kallsyms]                        [k] entry_SYSCALL_64_after_hwfr
+    3.50%     0.00%  CPU 0/KVM        [kernel.kallsyms]                        [k] do_syscall_64
+    3.50%     0.00%  CPU 0/KVM        [kernel.kallsyms]                        [k] __x64_sys_ioctl
+    3.50%     0.00%  CPU 0/KVM        [kernel.kallsyms]                        [k] kvm_vcpu_ioctl
+    3.48%     0.05%  CPU 0/KVM        [kernel.kallsyms]                        [k] kvm_arch_vcpu_ioctl_run
+    2.56%     0.07%  swapper          [kernel.kallsyms]                        [k] asm_sysvec_call_function_si
+    2.52%     0.02%  swapper          [kernel.kallsyms]                        [k] sysvec_call_function_single
+    2.41%     0.25%  swapper          [kernel.kallsyms]                        [k] __schedule
+    2.38%     0.07%  swapper          [kernel.kallsyms]                        [k] schedule_idle

 

[malabida]

The disk was dying. Replacing it fixed the problem