rgw/sfs: testing: TestSFSWALCheckpoint: drop explosion test to 300MB

Since adding storage_pool, the WAL doesn't explode quite as much as it used to with 10 threads and 1000 objects each (previously with multiple unpooled connections it'd reliably go over 500MB, but I've just seen a unit test where it "only" got to ~450MB, so let's drop the value we test against a bit such that we still confirm the problem, but have more wiggle room). Signed-off-by: Tim Serong <[email protected]>
aquarist-labs · Oct 25, 2023 · 6c7d871 · 6c7d871
1 parent 1e5aa11
commit 6c7d871
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/src/test/rgw/sfs/test_rgw_sfs_wal_checkpoint.cc b/src/test/rgw/sfs/test_rgw_sfs_wal_checkpoint.cc
@@ -114,9 +114,9 @@ TEST_F(TestSFSWALCheckpoint, confirm_wal_explosion) {
 
   // Using the SQLite default checkpointing mechanism with
   // 10 concurrent writer threads will easily push us past
-  // 500MB quite quickly.
+  // a few hundred megabytes quite quickly.
   std::uintmax_t max_wal_size = multithread_object_create(10, 1000);
-  EXPECT_GT(max_wal_size, SIZE_1MB * 500);
+  EXPECT_GT(max_wal_size, SIZE_1MB * 300);
 
   // The fact that we have no size limit set means the WAL
   // won't be truncated even when the last writer completes,