HOTFIX some bugs found during the installation

include/cloysterhpc/services/options.h

@@ -21,7 +21,7 @@ struct Options final {
     bool runAsDaemon;
     bool airGap;
     bool unattended;
-    bool disableMirrors;
+    bool enableMirrors;
     std::size_t logLevelInput;
     std::string error;
     std::string config;

include/cloysterhpc/services/runner.h

@@ -9,6 +9,7 @@
 #include <boost/process.hpp>
 #include <fmt/format.h>
 
+#include <stdexcept>
 #include <string>
 #include <vector>
 
@@ -17,35 +18,96 @@
 #include <cloysterhpc/services/options.h>
 #include <cloysterhpc/services/scriptbuilder.h>
 
-namespace cloyster::services::runner {
+namespace cloyster::services::runner::shell {
+
+namespace unsafe {
+    template <typename... Args>
+    [[nodiscard]]
+    int fmt(std::vector<std::string>& output,
+        fmt::format_string<Args...> format, Args&&... args)
+    {
+        auto command = fmt::format(format, std::forward<Args>(args)...);
+
+        auto opts = cloyster::Singleton<cloyster::services::Options>::get();
+        if (!opts->dryRun) {
+            LOG_DEBUG("Running shell command: {}", command);
+            boost::process::ipstream pipe_stream;
+            boost::process::child child("/bin/bash", "-c", command,
+                boost::process::std_out > pipe_stream);
+
+            std::string line;
+            while (pipe_stream && std::getline(pipe_stream, line)) {
+                output.emplace_back(line);
+                LOG_TRACE("{}", line);
+            }
+
+            child.wait();
+            LOG_DEBUG("Exit code: {}", child.exit_code());
+            return child.exit_code();
+        } else {
+            LOG_INFO("Dry Run: {}", command);
+            return 0;
+        }
+    }
 
-template <typename... Args>
-int shellfmt(fmt::format_string<Args...> fmt, Args&&... args)
-{
-    auto command = fmt::format(fmt, std::forward<Args>(args)...);
-
-    auto opts = cloyster::Singleton<cloyster::services::Options>::get();
-    if (!opts->dryRun) {
-        LOG_DEBUG("Running shell command: {}", command);
-        boost::process::ipstream pipe_stream;
-        boost::process::child child(
-            "/bin/bash", "-c", command, boost::process::std_out > pipe_stream);
-
-        std::string line;
-        while (pipe_stream && std::getline(pipe_stream, line)) {
-            LOG_TRACE("{}", line);
+    template <typename... Args>
+    [[nodiscard]]
+    int fmt(fmt::format_string<Args...> format, Args&&... args)
+    {
+        auto command = fmt::format(format, std::forward<Args>(args)...);
+
+        auto opts = cloyster::Singleton<cloyster::services::Options>::get();
+        if (!opts->dryRun) {
+            LOG_DEBUG("Running shell command: {}", command);
+            boost::process::ipstream pipe_stream;
+            boost::process::child child("/bin/bash", "-c", command,
+                boost::process::std_out > pipe_stream);
+
+            std::string line;
+            while (pipe_stream && std::getline(pipe_stream, line)) {
+                LOG_TRACE("{}", line);
+            }
+
+            child.wait();
+            LOG_DEBUG("Exit code: {}", child.exit_code());
+            return child.exit_code();
+        } else {
+            LOG_INFO("Dry Run: {}", command);
+            return 0;
         }
+    }
+}
 
-        child.wait();
-        LOG_DEBUG("Exit code: {}", child.exit_code());
-        return child.exit_code();
-    } else {
-        LOG_INFO("Dry Run: {}", command);
-        return 0;
+template <typename... Args>
+void fmt(fmt::format_string<Args...> format, Args&&... args)
+{
+    const std::string command
+        = fmt::format(format, std::forward<Args>(args)...);
+    const auto exitCode = unsafe::fmt("{}", command);
+    if (exitCode != 0) {
+        throw std::runtime_error(fmt::format(
+            "Command {} failed with exit code {}", command, exitCode));
     }
 }
 
-int shell(std::string_view cmd);
+void cmd(std::string_view cmd);
+
+template <typename... Args>
+[[nodiscard]]
+std::string output(fmt::format_string<Args...> format, Args&&... args)
+{
+    std::vector<std::string> output;
+    // Cosntruct a command here because it will be used twice and
+    // we can't use args twice without hurting the perfect forwarding
+    const std::string command
+        = fmt::format(format, std::forward<Args>(args)...);
+    const auto exitCode = unsafe::fmt(output, "{}", command);
+    if (exitCode != 0) {
+        throw std::runtime_error(fmt::format(
+            "Command {} failed with exit code {}", command, exitCode));
+    }
+    return fmt::format("{}", fmt::join(output, "\n"));
+}
 
 }
 

repos/repos.conf

@@ -122,8 +122,9 @@ filename=xcat-deps.repo
 upstream.repo=http://xcat.org/files/xcat/repos/yum/devel/xcat-dep/rh{releasever}/{arch}/
 upstream.gpgkey=http://xcat.org/files/xcat/repos/yum/devel/xcat-dep/rh{releasever}/{arch}/repodata/repomd.xml.key
 
+# ofedVersion is configured in the answerfile at ofed.version
 [doca]
 name=NVIDIA DOCA Repository - RHEL rhel{osversion}
 filename=mlx-doca.repo
-upstream.repo=https://linux.mellanox.com/public/repo/doca/latest/rhel{osversion}/{arch}/
+upstream.repo=https://linux.mellanox.com/public/repo/doca/{ofedVersion}/rhel{osversion}/{arch}/
 upstream.gpgkey=https://linux.mellanox.com/public/repo/doca/latest/rhel{osversion}/{arch}/GPG-KEY-Mellanox.pub

rpmspecs/opencattus.spec

@@ -1,6 +1,6 @@
 Name: opencattus-installer 
 Version: 1.0
-Release: 2
+Release: 4
 Summary: OpenCATTUS Installer
 License: Apache 2.0
 URL: https://versatushpc.com.br/opencattus/
@@ -49,6 +49,10 @@ install -m 644 repos/rocky-vault.conf %{buildroot}/opt/cloysterhpc/conf/repos/ro
 /opt/cloysterhpc/conf/repos/rocky-vault.conf
 
 %changelog
+* Thu Aug 14 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-4 - Bugfix 
+- Update OFED
+- Dump configuration
+- Add support for Rocky Linux 9.6
 * Wed Jul 16 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-3 - Add ansible roles
 - Add ansible roles implementation
 - Fix dnssec configuration generation in xCAT plugin

src/diskImage.cpp

@@ -38,30 +38,43 @@ void DiskImage::setPath(const std::filesystem::path& path)
 
 bool DiskImage::isKnownImage(const std::filesystem::path& path)
 {
+    constexpr auto chooseDistro = [](std::string_view imageView)
+        -> std::optional<cloyster::models::OS::Distro> {
+        if (imageView.starts_with("Rocky")) {
+            return cloyster::models::OS::Distro::Rocky;
+        } else if (imageView.starts_with("rhel")) {
+            return cloyster::models::OS::Distro::RHEL;
+        } else if (imageView.starts_with("OracleLinux")) {
+            return cloyster::models::OS::Distro::OL;
+        } else if (imageView.starts_with("AlmaLinux")) {
+            return cloyster::models::OS::Distro::AlmaLinux;
+        } else {
+            return std::nullopt;
+        }
+    };
+
     for (const auto& image : m_knownImageFilename) {
         if (path.filename().string() == image) {
             LOG_TRACE("Disk image is recognized")
 
             auto imageView = std::string_view(image);
-            if (imageView.starts_with("Rocky")) {
-                m_distro = cloyster::models::OS::Distro::Rocky;
-            } else if (imageView.starts_with("rhel")) {
-                m_distro = cloyster::models::OS::Distro::RHEL;
-            } else if (imageView.starts_with("OracleLinux")) {
-                m_distro = cloyster::models::OS::Distro::OL;
-            } else if (imageView.starts_with("AlmaLinux")) {
-                m_distro = cloyster::models::OS::Distro::AlmaLinux;
-            } else {
-                throw std::logic_error(fmt::format(
-                    "Can't determine the distro for the image {}", image));
+            const auto distro = chooseDistro(imageView);
+            if (distro) {
+                m_distro = distro;
+                return true;
             }
-
-            return true;
         }
     }
 
-    LOG_TRACE("Disk image is unknown. Maybe you're using a custom image or "
-              "changed the default name?");
+    const auto distro
+        = chooseDistro(std::string_view(path.filename().string()));
+    if (distro) {
+        m_distro = distro;
+        return true;
+    }
+    cloyster::functions::abort(
+        "Disk image is unknown. Maybe you're using a custom image or "
+        "changed the default name?");
     return false;
 }
 

src/models/answerfile.cpp

@@ -7,10 +7,12 @@
 #include <boost/algorithm/string/join.hpp>
 #include <boost/algorithm/string/split.hpp>
 #include <boost/lexical_cast.hpp>
+#include <chrono>
 #include <cloysterhpc/functions.h>
 #include <cloysterhpc/models/answerfile.h>
 #include <cloysterhpc/services/log.h>
 #include <cloysterhpc/services/options.h>
+#include <cloysterhpc/services/osservice.h>
 #include <cstddef>
 #include <fmt/core.h>
 #include <iterator>
@@ -426,7 +428,24 @@ void AnswerFile::loadSystemSettings()
     }
 
     system.version = m_keyfile.getString("system", "version");
-    system.kernel = m_keyfile.getString("system", "kernel");
+    const auto kernel = m_keyfile.getStringOpt("system", "kernel");
+    if (kernel) {
+        system.kernel = kernel.value();
+        LOG_INFO("Kernel override in the answerfile {}", system.kernel);
+    } else {
+        const auto latestKernelVersion = services::runner::shell::output(
+            // This runs very early so it stops loading all repositories caches,
+            // which is unecessary, so I pinned --repo=appstream here
+            "dnf list --repo=appstream kernel-devel --available "
+            "--showduplicates | sed 1d | "
+            // captures the kernel arch, e.g. x86_64, in $1
+            // this is important later when used to access /lib/modules/...
+            // folders
+            R"(perl -lane '$F[0] =~ s/kernel-devel\.(.*)$//; printf "%s.%s\n", $F[1], $1' | tail -1)");
+        system.kernel = latestKernelVersion;
+        LOG_INFO("Kernel omitted in the answerfile, using running kernel {}",
+            system.kernel);
+    }
 }
 
 AFNode AnswerFile::loadNode(const std::string& section)

src/ofed.cpp

@@ -3,6 +3,8 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 
+#include <fmt/core.h>
+
 #include <cloysterhpc/cloyster.h>
 #include <cloysterhpc/functions.h>
 #include <cloysterhpc/ofed.h>
@@ -47,6 +49,8 @@ bool OFED::installed() const
 void OFED::install() const
 {
     const auto opts = cloyster::Singleton<cloyster::services::Options>::get();
+    const auto cluster = cloyster::Singleton<cloyster::models::Cluster>::get();
+    const auto osinfo = cluster->getNodes()[0].getOS();
 
     if (opts->dryRun) {
         LOG_WARN("Dry-Run: Skiping OFED installation");
@@ -73,14 +77,16 @@ void OFED::install() const
                 cloyster::services::repos::RepoManager>::get();
             auto osService
                 = cloyster::Singleton<cloyster::services::IOSService>::get();
+            const std::string kernelVersion = std::string(osinfo.getKernel());
             repoManager->enable("doca");
             // Install the required packages
             runner->checkCommand("dnf makecache --repo=doca");
             runner->checkCommand(
-                "dnf -y install kernel kernel-devel doca-extra");
+                fmt::format("dnf -y install kernel-{kernelVersion} "
+                            "kernel-devel-{kernelVersion} doca-extra",
+                    fmt::arg("kernelVersion", kernelVersion)));
 
-            if (osService->getKernelRunning()
-                != osService->getKernelInstalled()) {
+            if (osService->getKernelRunning() != kernelVersion) {
                 LOG_WARN("New kernel installed! Rebooting after the "
                          "installation finishes is advised!");
             }
@@ -96,10 +102,9 @@ void OFED::install() const
             // The driver may support weak updates modules and load without
             // need for reboot.
             if (!opts->shouldSkip("compile-doca-driver")) {
-                runner->checkCommand(
-                    "bash -c \"/opt/mellanox/doca/tools/doca-kernel-support -k "
-                    "$(rpm -q --qf \"%{VERSION}-%{RELEASE}.%{ARCH}\n\" "
-                    "kernel-devel)\"");
+                runner->checkCommand(fmt::format(
+                    "/opt/mellanox/doca/tools/doca-kernel-support -k {}",
+                    kernelVersion));
             }
 
             // Get the last rpm in /tmp/DOCA*/ folder
@@ -123,4 +128,4 @@ void OFED::install() const
 
             break;
     }
-}
+}

src/services/ansible/roles/base.cpp

@@ -58,8 +58,18 @@ ScriptBuilder installScript(
 
     builder.addNewLine().addCommand("# Install general base packages");
 
-    std::set<std::string> allPackages = { "wget", "curl", "dnf-plugins-core",
-        "chkconfig", "jq", "tar", "python3-dnf-plugin-versionlock" };
+    // "python3-dnf-plugin-versionlock" is conflicting with dnf-plugins-core
+    // during the first install
+    std::set<std::string> allPackages = {
+        "wget",
+        "curl",
+        "dnf-plugins-core",
+        "chkconfig",
+        "initscripts", // @FIXME: This is only required if the provisioner is
+                       // xCAT
+        "jq",
+        "tar",
+    };
     if (const auto iter = role.m_vars.find("base_packages");
         iter != role.m_vars.end()) {
         for (const auto& pkg :

src/services/ansible/roles/spack.cpp

@@ -25,6 +25,8 @@ ScriptBuilder installScript(
         fmt::format("Expected spack role, found {}", role.m_roleName));
 
     builder.addNewLine()
+        .addCommand("# Exit early if spack is already installed")
+        .addCommand("test -d /opt/spack/.git && exit 0")
         .addCommand("# Install dependencies for Spack")
         .addPackage("git")
         .addNewLine()

src/services/ansible/roles/todo.txt

@@ -0,0 +1,10 @@
+Base (open)
+    1. confluent
+    2. RHEL10 sup
+
+Features (closed)
+    1. ood
+    2. grafana
+    3. slurm 
+    4. integração ood slurm
+    5. lustre

src/services/options.cpp

@@ -24,7 +24,7 @@ std::unique_ptr<Options> options::factory(int argc, const char** argv)
         .runAsDaemon = false,
         .airGap = false,
         .unattended = false,
-        .disableMirrors = false,
+        .enableMirrors = false,
         .logLevelInput = 3,
         .error = "NO ERROR",
         .config = "",
@@ -47,8 +47,7 @@ std::unique_ptr<Options> options::factory(int argc, const char** argv)
     app.add_flag("-t,--tui", opt.enableTUI, "Enable TUI");
     app.add_flag("-c,--cli", opt.enableCLI, "Enable CLI");
     app.add_flag("-D,--daemon", opt.runAsDaemon, "Run as daemon");
-    app.add_flag(
-        "--disable-mirrors", opt.disableMirrors, "Disable mirror URLs");
+    app.add_flag("--enable-mirrors", opt.enableMirrors, "Disable mirror URLs");
     app.add_option("--mirror-url", opt.mirrorBaseUrl, "Base URL for mirror")
         ->default_str("https://mirror.versatushpc.com.br");
     app.add_option(