Fix xCAT Installation

Author: dhilst

rpmspecs/opencattus.spec

@@ -1,6 +1,6 @@
 Name: opencattus-installer-debug
 Version: 1.0
-Release: 5
+Release: 6
 Summary: OpenCATTUS Installer
 License: Apache 2.0
 URL: https://versatushpc.com.br/opencattus/
@@ -49,7 +49,9 @@ install -m 644 repos/rocky-vault.conf %{buildroot}/opt/cloysterhpc/conf/repos/ro
 /opt/cloysterhpc/conf/repos/rocky-vault.conf
 
 %changelog
-* Thu Sep 16 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-6 - Support Confluent
+* Thu Sep 16 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-6 - xCAT bugfixes
+- Fix xCAT installation after adding Confluent
+* Thu Sep 16 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-5 - Support Confluent
 - Add support to the Lenovo HPC Confluent provisioner
 - Add --roles command line
 * Thu Aug 14 2025  Daniel Hilst <daniel@versatushpc.com.br> - 1.0-4 - Bugfix 

src/diskImage.cpp

@@ -13,6 +13,7 @@
 #include <cloysterhpc/services/options.h>
 #include <cloysterhpc/services/cache.h>
 #include <cloysterhpc/utils/singleton.h>
+#include <cloysterhpc/utils/optional.h>
 #include <unordered_map>
 
 // @FIXME: This file need some work
@@ -39,7 +40,7 @@ void DiskImage::setPath(const std::filesystem::path& path)
 
 bool DiskImage::isKnownImage(const std::filesystem::path& path)
 {
-    constexpr auto chooseDistro = [](std::string_view imageView)
+    constexpr auto toDistroImage = [](std::string_view imageView)
         -> std::optional<cloyster::models::OS::Distro> {
         if (imageView.starts_with("Rocky")) {
             return cloyster::models::OS::Distro::Rocky;
@@ -49,33 +50,24 @@ bool DiskImage::isKnownImage(const std::filesystem::path& path)
             return cloyster::models::OS::Distro::OL;
         } else if (imageView.starts_with("AlmaLinux")) {
             return cloyster::models::OS::Distro::AlmaLinux;
-        } else {
-            return std::nullopt;
         }
+        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);
-            const auto distro = chooseDistro(imageView);
-            if (distro) {
-                m_distro = distro;
+            m_distro = toDistroImage(imageView);
+            if (m_distro.has_value()) {
                 return true;
             }
         }
     }
-
-    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?");
+    m_distro = toDistroImage(path.filename().string());
+    LOG_TRACE("Disk image is unknown. Maybe you're using a custom image or "
+              "changed the default name?");
     return false;
 }
 

src/ofed.cpp

@@ -118,6 +118,7 @@ void OFED::install() const
             runner->checkCommand("dnf makecache --repo=doca*");
             runner->checkCommand("dnf install --nogpg -y doca-ofed mlnx-fw-updater");
             runner->executeCommand("systemctl restart openibd");
+            // runner->executeCommand("systemctl enable --now opensmd");
         } break;
 
         case OFED::Kind::Oracle:

src/services/ansible/roles/base.cpp

@@ -22,7 +22,6 @@ namespace cloyster::services::ansible::roles::base {
 ScriptBuilder installScript(
     const Role& role, const cloyster::models::OS& osinfo)
 {
-    LOG_DEBUG("Running base role");
     using namespace cloyster;
     ScriptBuilder builder(osinfo);
 
@@ -36,10 +35,15 @@ ScriptBuilder installScript(
         case models::OS::Distro::Rocky:
         case models::OS::Distro::AlmaLinux:
             LOG_DEBUG("Running base role");
-            builder.addPackage("epel-release");
             break;
 
         case models::OS::Distro::OL:
+            // @FIXME: This breaks the RepoManager logic. Package installing
+            //   repository files at /etc/yum.repos.d/, may install repositories
+            //   using metalink or mirrorlist, which triggers a bug in RepoManager
+            //   when the xCAT image is being generated. The RepoManager only
+            //   supports baseurl for now, metalink and mirror lists trigger a
+            //   bad optional access during runtime.
             switch (osinfo.getPlatform()) {
                 case models::OS::Platform::el8:
                     builder.addPackage("oracle-epel-release-el8");

src/services/ansible/roles/network.cpp

@@ -43,7 +43,7 @@ void configureManagementNetwork(const Connection& connection)
 {
     const std::string_view ipv6Method = cluster()->getProvisioner()
             == cloyster::models::Cluster::Provisioner::xCAT
-        ? "diabled"
+        ? "disabled"
         : "link-local";
 
     auto interface = connection.getInterface().value();
@@ -69,7 +69,7 @@ nmcli connection add type {type} mtu {mtu} ifname {iface} con-name {conn_name} \
     ipv4.ignore-auto-routes yes \
     ipv6.method {ipv6_method}
 sleep 0.2
-nmcli -w 10 device connect {iface}
+nmcli device connect {iface}
 )",
         fmt::arg("iface", interface),
         fmt::arg("conn_name", connectionName),
@@ -105,7 +105,7 @@ nmcli connection add type {type} mtu {mtu} ifname {iface} con-name {conn_name} \
     ipv4.ignore-auto-dns yes \
     ipv4.ignore-auto-routes yes \
 sleep 0.2
-nmcli -w 10 device connect {iface}
+nmcli device connect {iface}
 )",
         fmt::arg("iface", interface),
         fmt::arg("conn_name", connectionName),
@@ -140,7 +140,7 @@ nmcli connection add type {type} mtu {mtu} ifname {iface} con-name {conn_name} \
     ipv4.ignore-auto-dns yes \
     ipv4.ignore-auto-routes yes \
 sleep 0.2
-nmcli -w 10 device connect {iface}
+nmcli device connect {iface}
 )",
         fmt::arg("iface", interface),
         fmt::arg("conn_name", connectionName),
@@ -218,13 +218,15 @@ namespace cloyster::services::ansible::roles::network {
 
 void run(const Role& role)
 {
-    const auto connections = cluster()->getHeadnode().getConnections();
-    LOG_INFO("Setting up networks")
+    LOG_INFO("Setting up networks, use `--skip network` to skip")
+    if (utils::singleton::options()->shouldSkip("network")) {
+        return;
+    }
 
+    const auto connections = cluster()->getHeadnode().getConnections();
     configureNetworks(connections);
     configureFQDN();
     configureHostsFile();
-
 }
 
 }

src/services/ansible/roles/ofed.cpp

@@ -28,6 +28,10 @@ namespace cloyster::services::ansible::roles::ofed {
 
 void run(const Role& role)
 {
+    LOG_INFO("Setting up Infiniband, use `--skip infiniband` to skip");
+    if (cloyster::utils::singleton::options()->shouldSkip("infiniband")) {
+        return;
+    }
     configureInfiniband();
 }
 

src/services/ansible/roles/provisioner.cpp

@@ -24,7 +24,9 @@ void run(const Role& role)
     } else if  (provisioner == "xcat") {
         roles::run(Roles::XCAT, osinfo);
     } else {
-        cloyster::functions::abort("Expecing xcat or confluent at system.provisioner, found: {}", provisioner);
+        cloyster::functions::abort("Expecing xcat or confluent at "
+                                   "[system].provisioner, found: {}, at {}",
+            provisioner, utils::singleton::answerfile()->path());
     }
 }
 

src/services/ansible/roles/spack.cpp

@@ -21,6 +21,11 @@ ScriptBuilder installScript(
     using namespace cloyster;
     ScriptBuilder builder(osinfo);
 
+    LOG_INFO("Setting up Spack, use `--skip spack` to skip");
+    if (utils::singleton::options()->shouldSkip("spack")) {
+        return builder;
+    }
+
     LOG_ASSERT(role.roleName() == "spack",
         fmt::format("Expected spack role, found {}", role.roleName()));