Chain Reorganisation

A chain reorganisation (or "reorg") takes place when your node receives blocks that are part of a new longest chain. Your node will deactivate blocks in its old longest chain in favour of the blocks that build the new longest chain.

This process allows individual nodes across the network to agree on the same version of the blockchain, because the globally accepted view of the blockchain will always be the one with the longest* chain of blocks.

*Technically it's the chain with the most amount of work, but the most number of blocks is usually the same thing.

Scenario

When does a chain reorganisation take place?

A chain reorganisation most commonly takes place after two blocks have been mined at the same time.

Due to the propagation speed of blocks across the network, some nodes will receive the one block first, and some nodes will receive the other block first. Therefore, there will be a temporary disagreement about which of these blocks was actually "first" and belongs at the top of everyone's blockchain.

So how can we resolve this dispute and make sure everyone agrees upon the same version of the blockchain?

Well, it's resolved when the next block is mined. The next block to be mined will build on top of one of these blocks, creating a new longest chain. When nodes receive this newest block, they will see that it creates a new longest chain, and they will perform a chain reorganisation to adopt it.

So thanks to chain reorganisations, each node eventually agrees on the same version of the blockchain as everyone else.

Stale Blocks

What happens to the transactions in the old longest chain?

If a block is deactivated due to a chain reorganisation ("stale block"), the transactions inside it are no longer part of the blockchain.

So if you tried to spend the outputs from a transaction inside a stale block, nodes would reject your transaction because you are trying to spend bitcoins that do not exist in the valid chain.

Practically speaking though, if two blocks are mined at the same time, they're probably going to include the same (or similar) transactions in them, so a reorg isn't usually going to cause a problem.

However, if there are transactions in the stale block that are not in the competing block, they will get sent back in to your node's memory pool and propagated around the network again for the chance to be mined in to a future block.

But this is not a guarantee, and if a transaction does not exist in the active chain, it may as well have never happened.

It's worth waiting for a transaction to make it 2+ blocks in to the blockchain before considering it as final. There is always the chance that it will get reorg'd out, and you will have to wait/hope for it to be re-mined back in to the longest chain.

A "stale block" is sometimes referred to as an "orphan block", but "stale block" is the more accurate term. An orphan block is when your node receives a block before receiving the parent block that it builds upon, and stale blocks aren't orphans because they have parents.

Example

Here we can see an example of an actual chain reorganisation that took place in the blockchain at block height 578,141.

Size

How big can chain reorganisations be?

A chain reorganisation can be any number of blocks in length.

If your node receives a new chain of blocks that's longer than your current active chain, your node will do a chain reorganisation to adopt the new chain, no matter how many blocks will be replaced.

This is why it's possible for a miner with a majority of hashing power to replace blocks and transactions in your current longest chain via a 51% attack. The longest chain always wins.

However, "natural" chain reorganisations (ones that take place due to two blocks being mined at the same time) rarely involve more than the top block in your chain.

Frequency

How often do chain reorganisations happen?

Not very often. For your node to experience an honest chain reorganisation, the following needs to take place:

1. Two blocks are mined at the same time.
2. Your node receives one of the blocks first, but the other block ends up being mined on top of and becomes the new longest chain.

I don't know what the mathematical probability of this is, so here's the frequency of chain reorganisations based on the data from my bitcoin node (which has been running continuously since 17 December 2016):

Commands

How can you find chain reorganisations?

You can see the chain reorganisations your node has observed with the bitcoin-cli getchaintips command.

For example:

[
  {
    "height": 589919,
    "hash": "000000000000000000149b18e74316248d106e42ca410f509305ae58ccda6b13",
    "branchlen": 0,
    "status": "active"
  },
  {
    "height": 578141,
    "hash": "0000000000000000001253a5f37d3763dbe928d21f7d72a708f05268c044179c",
    "branchlen": 1,
    "status": "valid-fork"
  },
  {
    "height": 575695,
    "hash": "0000000000000000002409ed07fdbb1d0359a0c516014115c5451aea724baec8",
    "branchlen": 1,
    "status": "valid-headers"
  },
  ...

• brachlen – tells you how many blocks were in the competing chain of blocks.
• status – indicates the following:
  • active – This is our current active chain (the longest chain).
  • valid-fork – Our node performed a chain reorganisation. We downloaded and validated these blocks and had them as part of our active chain, but we later deactivated them after receiving a new longer chain of blocks.
  • valid-headers – Our node observed a possible chain reorganisation. We downloaded these blocks, but did not validate them as our active chain was equivalent and became longer.
  • headers-only – Our node observed a possible chain reorganisation. We received the block headers for a competing chain but did not download the full blocks.
  • invalid – A branch that contains invalid blocks.

The branches with the status of valid-fork are the ones with blocks that we originally considered part of our active blockchain, but we later deactivated them after receiving a new longer chain of blocks.

The branches with the status of valid-headers are competing chains that we received after already having an equivalent active chain. These could have resulted in a reorganisation, but our chain continued as the longest chain, so no reorganisation took place.

It's unlikely that you will see any chain reorganisations if you have not been running your node continuously for a few weeks or months. When your node downloads the blockchain it will download the blocks in the current longest chain only (and not blocks from any branches or old chain reorganisations). Your node needs to experience chain reorganisations as they happen for them to show up in bitcoin-cli getchaintips.

Summary

Chain reorganisations are a perfectly normal part of a Bitcoin node's function. Adopting the longest known chain allows nodes across a network to agree on the same blockchain, and chain reorganisations are just a part of this process.

Transactions inside blocks that are deactivated due to a chain reorganisation will become invalid, but they will be recycled in to the memory pool for the opportunity to be mined in to a block on the new longest chain.

So basically, if your transaction gets mined in to a block, there's still a chance that it could get thrown back in to the mempool due to a chain reorganisation. However, these natural reorgs typically only happen to the top block on the chain, so you should wait for your transaction to make it 2 blocks deep to help avoid this scenario.

Resources

• How to detect a fork with bitcoin-cli?
• Understanding getchaintips in terms of chain reorganisations
• The difference between stale and orphan blocks