migration: update cometbft tx size for t78 migration

[conorsch]

Describe your changes

We've already updated the CometBFT config template, but running nodes on the current testnet will still be using the prior value, generated at time of join. Let's update the value as part of pd migrate, logging a warning if we fail to do so.

Issue ticket number and link

Refs #4594, #4582.

Checklist before requesting a review

• If this code contains consensus-breaking changes, I have added the "consensus-breaking" label. Otherwise, I declare my belief that there are not consensus-breaking changes, for the following reason:

  relevant to consensus-breaking logic and how it's handled via migrations

[conorsch]

Encountered some errors while testing this. In short, the TendermintConfig struct is stricter about type validation than CometBFT is. For example, seed_mode = "false" is accepted by CometBFT, but the struct demands seed_mode = false. Worse, an indexer node will fail, due to the psql indexer not being supported by tendermint-rs (#3188). Let's log warnings in those instances, rather than allowing the error to percolate up to bailing out on the top-level pd migrate action. If that still proves problematic, we can fall back to asking node operators to make the change manually.

[dynst]

Suggested change

	/// Edit the node's CometBFT config file to set mempool.max_tx_bytes to the required value.
	/// This value will affect consensus, but the config setting is specified for CometBFT
	/// Edit the node's CometBFT config file to set mempool.max_txs_bytes to the required value.
	/// This value won't affect consensus, but the config setting is specified for CometBFT

[conorsch]

I'm happy with these changes: the cometbft munging will bail out before overwriting if it cannot interpret the on-disk cometbft file, logging a warning. That works well enough with the variety of munged configs I tested against.

[erwanor]

There are two max_tx_bytes parameters that we want to change:

• the first one is a mempool config parameter that controls the maximum size of a transaction to be included in the mempool/gossiped to other nodes
• the second is a consensus parameter that controls the maximum size of block data (including headers)

We want to change both values:

• the mempool value should be set to something around 30KB (x-ref pd: pick a smaller value for mempool.max_tx_bytes #4629)
• the consensus value should be set to 1MIB (x-ref #block size pd: configure cometbft with max_bytes to 1MiB #4597)

[dynst]

There are two different mempool configuration parameters that are easily mixed up.

• max_txs_bytes is what this PR is about
• max_tx_bytes is what pd: pick a smaller value for mempool.max_tx_bytes #4629 is about, though it'd make sense to add it to this PR, too, after template(comet): lower mempool.max_tx_bytes from 1MiB to 30KiB #4632 gets merged.

[conorsch]

The max_tx_bytes vs max_txs_bytes distinction is certainly ripe for confusion, but if we're changing two values in the CometBFT config for Testnet 78, then we should update both values as part of the migration flow. I'll make a change to this PR to accommodate.

[conorsch]

This is ready for review again. After migration, the cometbft config's mempool block should look like:

[mempool]
recheck = true
broadcast = true
wal_dir = ""
size = 5000
max_txs_bytes = 10485760
cache_size = 10000
keep-invalid-txs-in-cache = false
max_tx_bytes = 30720
max_batch_bytes = 0

I made sure to test with a customized config that set persistent_peers, and those changes were preserved post-migration.

[erwanor]

Looks great