backport free disk space bugfix from upstream rocksdb (#14353)

jsteemann · KVS85 · web-flow · commit 37e9b607a37c · 2021-07-28T13:46:42.000+03:00
Co-authored-by: Vadim &lt;vadim@arangodb.com&gt;
diff --git a/3rdParty/rocksdb/6.8/env/fs_posix.cc b/3rdParty/rocksdb/6.8/env/fs_posix.cc
@@ -832,7 +832,17 @@ class PosixFileSystem : public FileSystem {
       return IOError("While doing statvfs", fname, errno);
     }
 
-    *free_space = ((uint64_t)sbuf.f_bsize * sbuf.f_bfree);
+    // sbuf.bfree is total free space available to root
+    // sbuf.bavail is total free space available to unprivileged user
+    //  sbuf.bavail <= sbuf.bfree ... pick correct based upon effective user id
+    if (geteuid()) {
+      // non-zero user is unprivileged, or -1 if error.  take more conservative
+      // size
+      *free_space = ((uint64_t)sbuf.f_bsize * sbuf.f_bavail);
+    } else {
+      // root user can access all disk space
+      *free_space = ((uint64_t)sbuf.f_bsize * sbuf.f_bfree);
+    }
     return IOStatus::OK();
   }
 
diff --git a/CHANGELOG b/CHANGELOG
@@ -1,6 +1,11 @@
 v3.8.1 (XXXX-XX-XX)
 -------------------
 
+* Backport bugfix from upstream rocksdb repository for calculating the free disk
+  space for the database directory. Before the bugfix, rocksdb could
+  overestimate the amount of free space when the arangod process was run as
+  non-privileged users.
+
 * Fixed a problem with active failover, where a failover could take 5 mins
   because the follower was caught in a bad state during replication. This fixes
   BTS-425.
diff --git a/arangod/RocksDBEngine/RocksDBEngine.cpp b/arangod/RocksDBEngine/RocksDBEngine.cpp
@@ -480,6 +480,18 @@ void RocksDBEngine::start() {
     }
   }
 
+  uint64_t totalSpace;
+  uint64_t freeSpace;
+  if (TRI_GetDiskSpaceInfo(_path, totalSpace, freeSpace).ok() && totalSpace != 0) {
+    LOG_TOPIC("b71b9", DEBUG, arangodb::Logger::ENGINES)
+      << "total disk space for database directory mount: " 
+      << totalSpace << " bytes"
+      << ", free disk space for database directory mount: " 
+      << freeSpace << " bytes"
+      << " (" << (100.0 * double(freeSpace) / double(totalSpace)) << "% free)";
+  }
+
+
   // options imported set by RocksDBOptionFeature
   auto const& opts = server().getFeature<arangodb::RocksDBOptionFeature>();
 
diff --git a/lib/Basics/files.cpp b/lib/Basics/files.cpp
@@ -2643,8 +2643,18 @@ arangodb::Result TRI_GetDiskSpaceInfo(std::string const& path,
     TRI_set_errno(TRI_ERROR_SYS_ERROR);
     return {TRI_errno(), TRI_last_error()};
   }
-  totalSpace = static_cast<uint64_t>(stat.f_frsize) * static_cast<uint64_t>(stat.f_blocks);
-  freeSpace = static_cast<uint64_t>(stat.f_frsize) * static_cast<uint64_t>(stat.f_bfree);
+  totalSpace = static_cast<uint64_t>(stat.f_bsize) * static_cast<uint64_t>(stat.f_blocks);
+
+  // sbuf.bfree is total free space available to root
+  // sbuf.bavail is total free space available to unprivileged user
+  // sbuf.bavail <= sbuf.bfree ... pick correct based upon effective user id
+  if (geteuid()) {
+    // non-zero user is unprivileged, or -1 if error. take more conservative size
+    freeSpace = static_cast<uint64_t>(stat.f_bsize) * static_cast<uint64_t>(stat.f_bavail);
+  } else {
+    // root user can access all disk space
+    freeSpace = static_cast<uint64_t>(stat.f_bsize) * static_cast<uint64_t>(stat.f_bfree);
+  }
 #endif
   return {};
 }
