clean cache slog files on startup (#143)

- Bugfix: If we fail to move a temporary file into place, try to re-open the file so we clean it up
- Improvement: Clean up cache bucket log files at startup as we store the matching information in the index snapshot for the bucket

1 files changed, 31 insertions, 19 deletions

diff --git a/src/zenstore/cache/cachedisklayer.cpp b/src/zenstore/cache/cachedisklayer.cpp
index 9c350b11d..37404053e 100644
--- a/src/zenstore/cache/cachedisklayer.cpp
+++ b/src/zenstore/cache/cachedisklayer.cpp
@@ -257,7 +257,7 @@ public:
 	Oid GenerateNewManifest(std::filesystem::path ManifestPath);
 
 	IoBuffer MakeSidecarManifest(const Oid& BucketId, uint64_t EntryCount);
-	uint64_t GetSidecarSize() const { return m_ManifestEntryCount * sizeof(ManifestData); }
+	uint64_t GetSidecarSize() const { return sizeof(BucketMetaHeader) + m_ManifestEntryCount * sizeof(ManifestData); }
 	void	 WriteSidecarFile(RwLock::SharedLockScope&											 BucketLock,
 							  const std::filesystem::path&										 SidecarPath,
 							  uint64_t															 SnapshotLogPosition,
@@ -792,11 +792,11 @@ ZenCacheDiskLayer::CacheBucket::OpenOrCreate(std::filesystem::path BucketDir, bo
 }
 
 void
-ZenCacheDiskLayer::CacheBucket::WriteIndexSnapshotLocked(const std::function<uint64_t()>& ClaimDiskReserveFunc)
+ZenCacheDiskLayer::CacheBucket::WriteIndexSnapshotLocked(bool FlushLockPosition, const std::function<uint64_t()>& ClaimDiskReserveFunc)
 {
 	ZEN_TRACE_CPU("Z$::Bucket::WriteIndexSnapshot");
 
-	const uint64_t LogCount = m_SlogFile.GetLogCount();
+	const uint64_t LogCount = FlushLockPosition ? 0 : m_SlogFile.GetLogCount();
 	if (m_LogFlushPosition == LogCount)
 	{
 		return;
@@ -880,19 +880,31 @@ ZenCacheDiskLayer::CacheBucket::WriteIndexSnapshotLocked(const std::function<uin
 		{
 			std::filesystem::path TempFilePath = ObjectIndexFile.GetPath();
 			ZEN_WARN("snapshot failed to rename new snapshot '{}' to '{}', reason: '{}'", TempFilePath, IndexPath, Ec.message());
-
-			if (std::filesystem::is_regular_file(TempFilePath))
-			{
-				if (!std::filesystem::remove(TempFilePath, Ec) || Ec)
-				{
-					ZEN_WARN("snapshot failed to remove temporary file {}, reason: '{}'", TempFilePath, Ec.message());
-				}
-			}
 		}
 		else
 		{
 			// We must only update the log flush position once the snapshot write succeeds
-			m_LogFlushPosition = LogCount;
+			if (FlushLockPosition)
+			{
+				std::filesystem::path LogPath = GetLogPath(m_BucketDir, m_BucketName);
+
+				if (std::filesystem::is_regular_file(LogPath))
+				{
+					if (!std::filesystem::remove(LogPath, Ec) || Ec)
+					{
+						ZEN_WARN("snapshot failed to clean log file '{}', removing index at '{}', reason: '{}'",
+								 LogPath,
+								 IndexPath,
+								 Ec.message());
+						std::error_code RemoveIndexEc;
+						std::filesystem::remove(IndexPath, RemoveIndexEc);
+					}
+				}
+			}
+			if (!Ec)
+			{
+				m_LogFlushPosition = LogCount;
+			}
 		}
 	}
 	catch (const std::exception& Err)
@@ -1104,6 +1116,11 @@ ZenCacheDiskLayer::CacheBucket::InitializeIndexFromDisk(RwLock::ExclusiveLockSco
 		}
 	}
 
+	if (IsNew || LogEntryCount > 0 || m_LogFlushPosition != 0)
+	{
+		WriteIndexSnapshot(IndexLock, /*Flush log*/ true);
+	}
+
 	m_SlogFile.Open(LogPath, CasLogFile::Mode::kWrite);
 
 	BlockStore::BlockIndexSet KnownBlocks;
@@ -1124,11 +1141,6 @@ ZenCacheDiskLayer::CacheBucket::InitializeIndexFromDisk(RwLock::ExclusiveLockSco
 		}
 	}
 	m_BlockStore.SyncExistingBlocksOnDisk(KnownBlocks);
-
-	if (IsNew || LogEntryCount > 0)
-	{
-		WriteIndexSnapshot(IndexLock);
-	}
 }
 
 void
@@ -1833,7 +1845,7 @@ ZenCacheDiskLayer::CacheBucket::SaveSnapshot(const std::function<uint64_t()>& Cl
 
 			{
 				RwLock::SharedLockScope IndexLock(m_IndexLock);
-				WriteIndexSnapshot(IndexLock);
+				WriteIndexSnapshot(IndexLock, /*Flush log*/ false);
 				// Note: this copy could be eliminated on shutdown to
 				// reduce memory usage and execution time
 				Index		= m_Index;
@@ -1873,7 +1885,7 @@ ZenCacheDiskLayer::CacheBucket::SaveSnapshot(const std::function<uint64_t()>& Cl
 		else
 		{
 			RwLock::SharedLockScope IndexLock(m_IndexLock);
-			WriteIndexSnapshot(IndexLock);
+			WriteIndexSnapshot(IndexLock, /*Flush log*/ false);
 			const uint64_t EntryCount = m_Index.size();
 			Buffer					  = ManifestWriter.MakeSidecarManifest(m_BucketId, EntryCount);
 			uint64_t SidecarSize	  = ManifestWriter.GetSidecarSize();