commit b6e2e0bff1098088cdc5daf79a6c9baeb6312dfc
author Jason Zhou <jasonzhou460@gmail.com> Wed Aug 07 14:44:12 2024 -0400
committer Benjamin Mahler <bmahler@apache.org> Wed Aug 07 14:46:15 2024 -0400
tree b5f20cb2024b5b9791467a139480b5258c44eef6
parent 043c6aa8c76c997354a6a46cb0823db6dcbde91e

[device manager] Checkpoint state on device manager state change.

Currently we do not checkpoint the device access state of each cgroup
when the configure or reconfigure is called. Meaning that we have
no way of recovering a cgroup's device access state.

We will checkpoint state of the device manager whenever its state
is being changed to ensure that we can recover the most recent state
when necessary.

Review: https://reviews.apache.org/r/75143/

src/slave/containerizer/device_manager/device_manager.cpp

diff --git a/src/slave/containerizer/device_manager/device_manager.cpp b/src/slave/containerizer/device_manager/device_manager.cpp
index e613323..a2b6b9c 100644
--- a/src/slave/containerizer/device_manager/device_manager.cpp
+++ b/src/slave/containerizer/device_manager/device_manager.cpp
@@ -27,7 +27,9 @@
 #include <stout/unreachable.hpp>
 
 #include "slave/containerizer/device_manager/device_manager.hpp"
+#include "slave/containerizer/device_manager/state.hpp"
 #include "slave/paths.hpp"
+#include "slave/state.hpp"
 #include "linux/cgroups2.hpp"
 
 using std::string;
@@ -227,10 +229,43 @@
 
   hashmap<string, DeviceManager::CgroupDeviceAccess> device_access_per_cgroup;
 
-  // TODO(jasonzhou): persist device_access_per_cgroup on disk.
+  Try<Nothing> checkpoint() const
+  {
+    CgroupDeviceAccessStates states;
+
+    foreachpair (const string& cgroup,
+                 const DeviceManager::CgroupDeviceAccess& access,
+                 device_access_per_cgroup) {
+      CgroupDeviceAccessState* state = &(*(states.mutable_device_access_per_cgroup()))[cgroup];
+
+      foreach (const Entry& entry, access.allow_list) {
+        state->add_allow_list(stringify(entry));
+      }
+      foreach (const Entry& entry, access.deny_list) {
+        state->add_deny_list(stringify(entry));
+      }
+    }
+
+    Try<Nothing> status =
+      state::checkpoint(paths::getDevicesStatePath(meta_dir), states);
+
+    if (status.isError()) {
+      return Error("Failed to perform checkpoint: " + status.error());
+    }
+
+    return Nothing();
+  }
+
   Try<Nothing> commit_device_access_changes(const string& cgroup) const
   {
-    Try<Nothing> status = cgroups2::devices::configure(
+    Try<Nothing> status = checkpoint();
+
+    if (status.isError()) {
+      return Error("Failed to checkpoint device access state: "
+                   + status.error());
+    }
+
+    status = cgroups2::devices::configure(
         cgroup,
         device_access_per_cgroup.at(cgroup).allow_list,
         device_access_per_cgroup.at(cgroup).deny_list);