PeerDAS internal availability check failure

[jimmygchen]

We're seeing the following errors on on a peerdas-devnet-5 node lighthouse-geth-1 :

Apr 04 09:43:49.680 WARN Internal availability check failure           block_root: 0xc98c2fca6e57d239fd4ad65cd09a8113f1e85608758c0303
89d1914b671b7e41, error: Unexpected("too many columns got 128 expected 0"), service: "lookup_sync"

This is a supernode, but its expecting 0 columns in lookup sync. I think it's because RpcBlock from lookup is not coupled with columns when we instantiate an RpcBlock with block variant here:

lighthouse/beacon_node/network/src/sync/block_lookups/common.rs

Line 103 in 7a0388e

RpcBlock::new_without_blobs(Some(block_root), value),

This could also happen via the HTTP duplicate block import path, or whenever we call new_without_blobs where blobs are expected:

lighthouse/beacon_node/beacon_chain/src/block_verification.rs

Line 1277 in 0e6da0f

.verify_kzg_for_rpc_block(RpcBlock::new_without_blobs(Some(block_root), self.clone()))

In these scenarios, the node would not be able to import the block.

Related PR:

• Remove CGC from data_availability checker #7033

[jimmygchen]

Fixed in #7281