fixed crash while listing tablet boot info backup that was corrupted; added replying with raw protobuf on list requests (#4565)

если мы не успели дописать бэкап следующим образом
```
      Channels {
        Channel: 5
        ChannelType: 0
        History {
          FromGeneration: 0
          GroupID: 2181038117
        }
        StoragePool: "/pre-prod_sas/NBS:rot"
      }
      Channels {
  ```

и пытаемся после этого подняться и прогреть группы, то протобуфф успешно парсится, но падает на верифайках при попытке почитать бэкап при конвертации из протобуфины во внутреннюю структуру.

Вызов функции с верифайками:
https://github.com/ydb-platform/nbs/blob/fdde68d77b9eb3b13e9c5a0fe8082599e2e30f63/cloud/storage/core/libs/hive_proxy/tablet_boot_info_backup.cpp#L272-L274
 https://github.com/ydb-platform/nbs/blob/fdde68d77b9eb3b13e9c5a0fe8082599e2e30f63/contrib/ydb/core/base/tablet.cpp#L5

Author: vladstepanyuk

cloud/blockstore/libs/daemon/ydb/bootstrap.cpp

@@ -241,17 +241,20 @@ class TWarmupBSGroupConnectionsActor final
         auto tabletBootInfos = std::move(ev->Get()->TabletBootInfos);
         THashSet<ui64> groupIds;
         for (const auto& tabletBootInfo: tabletBootInfos) {
-            for (const auto& channel: tabletBootInfo.StorageInfo->Channels) {
+            for (const auto& channel:
+                 tabletBootInfo.StorageInfoProto.GetChannels())
+            {
                 auto historyEntries =
-                    channel.History | std::views::reverse |
+                    channel.GetHistory() | std::views::reverse |
                     std::views::filter(
                         [&](const auto& el)
-                        { return groupIds.insert(el.GroupID).second; }) |
+                        { return groupIds.insert(el.GetGroupID()).second; }) |
                     std::views::take(GroupsPerChannelToWarmup);
                 for (const auto& historyEntry: historyEntries) {
                     NCloud::Send(
                         ctx,
-                        NKikimr::MakeBlobStorageProxyID(historyEntry.GroupID),
+                        NKikimr::MakeBlobStorageProxyID(
+                            historyEntry.GetGroupID()),
                         std::make_unique<NKikimr::TEvBlobStorage::TEvStatus>(
                             TInstant::Max()));
                 }

cloud/storage/core/libs/hive_proxy/hive_proxy_ut.cpp

@@ -2,6 +2,7 @@
 #include "hive_proxy_events_private.h"
 
 #include <cloud/storage/core/libs/api/hive_proxy.h>
+#include <cloud/storage/core/libs/hive_proxy/protos/tablet_boot_info_backup.pb.h>
 
 #include <contrib/ydb/core/base/blobstorage.h>
 #include <contrib/ydb/core/base/hive.h>
@@ -13,6 +14,7 @@
 #include <contrib/ydb/core/testlib/basics/helpers.h>
 #include <contrib/ydb/core/testlib/tablet_helpers.h>
 
+#include <library/cpp/protobuf/util/pb_io.h>
 #include <library/cpp/testing/unittest/registar.h>
 
 namespace NCloud::NStorage {
@@ -1672,19 +1674,86 @@ Y_UNIT_TEST_SUITE(THiveProxyTest)
             UNIT_ASSERT_VALUES_EQUAL(1, result.TabletBootInfos.size());
             UNIT_ASSERT_VALUES_EQUAL(
                 FakeTablet2,
-                result.TabletBootInfos[0].StorageInfo->TabletID);
+                result.TabletBootInfos[0].StorageInfoProto.GetTabletID());
             TVector<TVector<ui64>> actualGroupIds;
-            for (auto& channel: result.TabletBootInfos[0].StorageInfo->Channels)
+            for (auto& channel:
+                 result.TabletBootInfos[0].StorageInfoProto.GetChannels())
             {
                 actualGroupIds.emplace_back();
-                for (auto& historyEntry: channel.History) {
-                    actualGroupIds.back().emplace_back(historyEntry.GroupID);
+                for (auto& historyEntry: channel.GetHistory()) {
+                    actualGroupIds.back().emplace_back(
+                        historyEntry.GetGroupID());
                 }
             }
 
             UNIT_ASSERT_VALUES_EQUAL(expectedGroupIds, actualGroupIds);
         }
     }
+
+    Y_UNIT_TEST(ShouldListCorruptedTabletBootInfoBackupsWithoutCrash)
+    {
+        TString backupFilePath =
+            "BootExternal.tablet_boot_info_backup.txt";
+
+        NKikimr::TTabletStorageInfoPtr storageInfo;
+
+        TVector<TVector<ui64>> expectedGroupIds;
+        {
+            TTestBasicRuntime runtime;
+            TTestEnv env(runtime, backupFilePath, false);
+
+            TTabletStorageInfoPtr expected = CreateTestTabletInfo(
+                FakeTablet2,
+                TTabletTypes::BlockStorePartition);
+            env.HiveState->StorageInfos[FakeTablet2] = expected;
+
+            auto sender = runtime.AllocateEdgeActor();
+
+            auto result =
+                env.SendBootExternalRequest(sender, FakeTablet2, S_OK);
+            UNIT_ASSERT(result.StorageInfo);
+            UNIT_ASSERT_VALUES_EQUAL(FakeTablet2, result.StorageInfo->TabletID);
+            UNIT_ASSERT_VALUES_EQUAL(1u, result.SuggestedGeneration);
+
+            // Smoke check for background sync (15 seconds should be enough).
+            runtime.AdvanceCurrentTime(TDuration::Seconds(15));
+            runtime.DispatchEvents(TDispatchOptions(), TDuration::Seconds(15));
+
+            for (auto& channel: result.StorageInfo->Channels) {
+                expectedGroupIds.emplace_back();
+                for (auto& historyEntry: channel.History) {
+                    expectedGroupIds.back().emplace_back(
+                        historyEntry.GroupID);
+                }
+            }
+            storageInfo = result.StorageInfo;
+            env.SendBackupTabletBootInfos(sender, S_OK);
+        }
+
+        // simulate backup corruption
+        {
+            NHiveProxy::NProto::TTabletBootInfoBackup backupProto;
+            MergeFromTextFormat(backupFilePath, backupProto);
+
+            backupProto.MutableData()
+                ->begin()
+                ->second.MutableStorageInfo()
+                ->MutableChannels(0)
+                ->SetChannelErasureName("");
+
+            TFileOutput output(backupFilePath);
+            SerializeToTextFormat(backupProto, output);
+        }
+
+        {
+            TTestBasicRuntime runtime;
+            TTestEnv env(runtime, backupFilePath, false);
+
+            auto sender = runtime.AllocateEdgeActor();
+
+            env.SendListTabletBootInfoBackups(sender, S_OK);
+        }
+    }
 }
 
 }   // namespace NCloud::NStorage

cloud/storage/core/libs/hive_proxy/tablet_boot_info.h

@@ -4,6 +4,8 @@
 
 #include <cloud/storage/core/libs/kikimr/public.h>
 
+#include <contrib/ydb/core/protos/tablet.pb.h>
+
 namespace NCloud::NStorage {
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -13,13 +15,13 @@ struct TTabletBootInfo
     TTabletBootInfo() = default;
 
     TTabletBootInfo(
-            NKikimr::TTabletStorageInfoPtr storageInfo,
+            NKikimrTabletBase::TTabletStorageInfo storageInfoProto,
             ui64 suggestedGeneration)
-        : StorageInfo(std::move(storageInfo))
+        : StorageInfoProto(std::move(storageInfoProto))
         , SuggestedGeneration(suggestedGeneration)
     {}
 
-    NKikimr::TTabletStorageInfoPtr StorageInfo;
+    NKikimrTabletBase::TTabletStorageInfo StorageInfoProto;
     ui64 SuggestedGeneration = 0;
 };
 

cloud/storage/core/libs/hive_proxy/tablet_boot_info_backup.cpp

@@ -270,7 +270,7 @@ void TTabletBootInfoBackup::HandleListTabletBootInfoBackups(
     TVector<TTabletBootInfo> tabletBootInfos;
     for (const auto& [_, tabletBootInfo]: BackupProto.GetData()) {
         tabletBootInfos.emplace_back(
-            NKikimr::TabletStorageInfoFromProto(tabletBootInfo.GetStorageInfo()),
+            tabletBootInfo.GetStorageInfo(),
             tabletBootInfo.GetSuggestedGeneration());
     }