From dc0b78cdb88e9cbfd7dc1a2ee0b15333939be549 Mon Sep 17 00:00:00 2001 From: Alex Auvolat Date: Tue, 19 Mar 2024 11:04:20 +0100 Subject: [block-ref-repair] Block refcount recalculation and repair - We always recalculate the reference count of a block before deleting it locally, to make sure that it is indeed zero. - If we had to fetch a remote block but we were not able to get it, check that refcount is indeed > 0. - Repair procedure that checks everything --- src/block/resync.rs | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'src/block/resync.rs') diff --git a/src/block/resync.rs b/src/block/resync.rs index 48c2cef1..b4108213 100644 --- a/src/block/resync.rs +++ b/src/block/resync.rs @@ -367,6 +367,13 @@ impl BlockResyncManager { } if exists && rc.is_deletable() { + if manager.rc.recalculate_rc(hash)?.0 > 0 { + return Err(Error::Message(format!( + "Refcount for block {:?} was inconsistent, retrying later", + hash + ))); + } + info!("Resync block {:?}: offloading and deleting", hash); let existing_path = existing_path.unwrap(); @@ -453,7 +460,15 @@ impl BlockResyncManager { hash ); - let block_data = manager.rpc_get_raw_block(hash, None).await?; + let block_data = manager.rpc_get_raw_block(hash, None).await; + if matches!(block_data, Err(Error::MissingBlock(_))) { + warn!( + "Could not fetch needed block {:?}, no node returned valid data. Checking that refcount is correct.", + hash + ); + manager.rc.recalculate_rc(hash)?; + } + let block_data = block_data?; manager.metrics.resync_recv_counter.add(1); -- cgit v1.2.3