From ba352c01f659ff3f0b667db3cf7136a59658e0b9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Vin=C3=ADcius=20Ferr=C3=A3o?= <2031761+viniciusferrao@users.noreply.github.com> Date: Fri, 21 Aug 2026 00:05:48 -0300 Subject: [PATCH] docs(riscv64): describe crash dumps on riscv64 nodes Explain why the riscv64 templates disable the installer's kdump add-on, how to reserve memory for crash dumps on an installed node, and what diskless images reserve by default. --- .../manage_clusters/riscv64/index.rst | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/docs/source/guides/admin-guides/manage_clusters/riscv64/index.rst b/docs/source/guides/admin-guides/manage_clusters/riscv64/index.rst index f87a58381..ca16832bf 100644 --- a/docs/source/guides/admin-guides/manage_clusters/riscv64/index.rst +++ b/docs/source/guides/admin-guides/manage_clusters/riscv64/index.rst @@ -75,6 +75,22 @@ package lists add ``grub2-efi-riscv64`` and ``efibootmgr``, and the these files. After ``nodeset boot`` the firmware boots the installed system because the per-node ``grub2-`` loader link is removed. +Crash dumps +~~~~~~~~~~~ + +EL10 defines no default crash kernel reservation for riscv64 +(``kdumpctl get-default-crashkernel`` is empty there), so the installer's kdump +add-on writes the literal ``crashkernel=auto``, which the kernel ignores: no +memory is reserved and ``kdump.service`` fails on every installed node. The +riscv64 templates therefore turn that add-on off. To take crash dumps on a +riscv64 node, reserve the memory yourself with a real value, for example:: + + chdef -t osimage rocky10.2-riscv64-install-compute addkcmdline="crashkernel=256M" + +Diskless images use ``linuximage.dump`` and ``linuximage.crashkernelsize`` as on +the other architectures; a riscv64 image with ``dump`` set and no explicit size +reserves 256M. + Stateless (diskless) images ---------------------------