Skip to content

v1.26.2

Compare
Choose a tag to compare
@rvagg rvagg released this 08 Apr 23:13
· 705 commits to releases since this release
v1.26.2

This is a mandatory patch release for the Filecoin network version 22 mainnet upgrade, for all node operators.

There is an update in the upgrade epoch for nv22, you can read the full discussion in Slack here.

The new upgrade epoch is scheduled to be on epoch 3855360 - 2024-04-24 - 14:00:00Z. That means:

  • All mainnet node operators that have upgraded to v1.26.x, must upgrade to this patch release before 2024-04-11T14:00:00Z.
  • All mainnet node operators that are on a version lower the v1.26.x, must upgrade to this patch release before 2024-04-24T14:00:00Z.

Please note that releases v1.26.0 and v1.26.1 have been retracted and should not be used prior to the new network version 22 upgrade epoch.

This patch also includes fixes for node operators who want to index builtin-actor events after the nv22 upgrade. Specifically, it ensures the builtin actor event entries are ordered by insertion order when selected (#11834). It also includes a couple Lotus-Miner patch fixes, ensuring that SnapDeals works properly and are using the new ProveReplicaUpdate3 message after the network version 22 upgrade, ensuring that DDO-sectors has the correct sector expirations, as well as DDO-sector visibility in the lotus-miner sectors list cmd.

The information below is copied from the v1.26.0 release notes and relates to inclusions in the v1.26 series:


This is the stable release for the upcoming MANDATORY Filecoin network upgrade v22, codenamed Dragon 🐉, at epoch 3817920 - 2024-04-11 - 14:00:00Z epoch 3855360 - 2024-04-24 - 14:00:00Z The switch to the new Drand-network will happen exactly 120 epochs (1 hour) after the Dragon-upgrade. This switch has been codenamed Phoenix.

The Filecoin network version 22 delivers the following FIPs:

☢️ Upgrade Warnings ☢️

  • This release requires a minimum Go version of v1.21.7 or higher to successfully build Lotus.
  • SPs: Some SPs has encountered a wdpost with the address didn't have enough funds to send message error error when upgrading to this version. Due to #9746 which was introduced in Lotus v1.25.2, SPs must ensure they have a higher FIL amount in their windowPoSt-wallet then the MaxWindowPoStGasFee setting is. Alternetively they can set the MaximizeWindowPoStFeeCap settting to false.

v13 Builtin Actor Bundle

Builtin actor v13.0.0 is used for supporting this upgrade. Make sure that your lotus actor bundle matches the v13 actors manifest by running the following cli after upgrading:

lotus state actor-cids --network-version=22
Network Version: 22
Actor Version: 13
Manifest CID: bafy2bzacecdhvfmtirtojwhw2tyciu4jkbpsbk5g53oe24br27oy62sn4dc4e

Actor             CID  
account           bafk2bzacedxnbtlsqdk76fsfmnhyvsblwyfducerwwtp3mqtx2wbrvs5idl52
cron              bafk2bzacebbopddyn5csb3fsuhh2an4ttd23x6qnwixgohlirj5ahtcudphyc
datacap           bafk2bzaceah42tfnhd7xnztawgf46gbvc3m2gudoxshlba2ucmmo2vy67t7ci
eam               bafk2bzaceb23bhvvcjsth7cn7vp3gbaphrutsaz7v6hkls3ogotzs4bnhm4mk
ethaccount        bafk2bzaceautge6zhuy6jbj3uldwoxwhpywuon6z3xfvmdbzpbdribc6zzmei
evm               bafk2bzacedq6v2lyuhgywhlllwmudfj2zufzcauxcsvvd34m2ek5xr55mvh2q
init              bafk2bzacedr4xacm3fts4vilyeiacjr2hpmwzclyzulbdo24lrfxbtau2wbai
multisig          bafk2bzacecr5zqarfqak42xqcfeulsxlavcltawsx2fvc7zsjtby6ti4b3wqc
paymentchannel    bafk2bzacebntdhfmyc24e7tm52ggx5tnw4i3hrr3jmllsepv3mibez4hywsa2
placeholder       bafk2bzacedfvut2myeleyq67fljcrw4kkmn5pb5dpyozovj7jpoez5irnc3ro
reward            bafk2bzacedq4q2kwkruu4xm7rkyygumlbw2yt4nimna2ivea4qarvtkohnuwu
storagemarket     bafk2bzacebjtoltdviyznpj34hh5qp6u257jnnbjole5rhqfixm7ug3epvrfu
storageminer      bafk2bzacebf4rrqyk7gcfggggul6nfpzay7f2ordnkwm7z2wcf4mq6r7i77t2
storagepower      bafk2bzacecjy4dkulvxppg3ocbmeixe2wgg6yxoyjxrm4ko2fm3uhpvfvam6e
system            bafk2bzacecyf523quuq2kdjfdvyty446z2ounmamtgtgeqnr3ynlu5cqrlt6e
verifiedregistry  bafk2bzacedkxehp7y7iyukbcje3wbpqcvufisos6exatkanyrbotoecdkrbta

Migration

We are expecting a bit heavier than normal state migration for this upgrade due to the amount of state changes introduced with Direct Data Onboarding.

All node operators, including storage providers, should be aware that ONE pre-migration is being scheduled 120 epochs before the upgrade. It will take around 10-20 minutes for the pre-migration and less than 30 seconds for the final migration, depending on the amount of historical state in the node blockstore and the hardware specs the node is running on. During this time, expect slower block validation times, increased CPU and memory usage, and longer delays for API queries

We recommend node operators (who haven't enabled splitstore discard mode) that do not care about historical chain states, to prune the chain blockstore by syncing from a snapshot 1-2 days before the upgrade.

You can test out the migration by running running the benchmarking a network migration tutorial.

For certain node operators, such as full archival nodes or systems that need to keep large amounts of state (RPC providers), completing the pre-migration in time before the network upgrade might not be achievable. For those node operators, it is recommended to skip the pre-migration and run the non-cached migration (i.e., just running the migration at the exact upgrade epoch), and schedule for some downtime during the upgrade epoch. Operators of such nodes can read the How to disable premigration in network upgrade tutorial.

New features

Tracing API

Replace the CodeCid field in the message trace (added in 1.23.4) with an InvokedActor field.

Before:

{
    "Msg": {
        "From": ...,
        "To": ...,
        ...
        "CodeCid": ... // The actor's code CID.
    }
    "MsgRct": ...,
    "GasCharges": [],
    "Subcalls": [],
}

After:

{
    "Msg": {
        "From": ...,
        "To": ...
    }
    "InvokedActor": {         // The invoked actor (ommitted if the actor wasn't invoked).
        "Id": 1234,           // The ID of the actor.
        "State": {            // The actor's state object (may change between network versions).
           "Code": ...,       // The actor's code CID.
           "Head": ...,       // The actor's state-root (when invoked).
           "CallSeqNum": ..., // The actor's nonce.
           "Balance": ...,    // The actor's balance (when invoked).
           "Address": ...,    // Delegated address (FEVM only).
        }
    }
    "MsgRct": ...,
    "GasCharges": [],
    "Subcalls": [],
}

This means the trace now contains an accurate "snapshot" of the actor at the time of the call, information that may not be present in the final state-tree (e.g., due to reverts). This will hopefully improve the performance and accuracy of indexing services.

Ethereum Tracing API (trace_block and trace_replayBlockTransactions)

For those with the Ethereum JSON-RPC API enabled, the experimental Ethereum Tracing API has been improved significantly and should be considered "functional". However, it's still new and should be tested extensively before relying on it. This API translates FVM traces to Ethereum-style traces, implementing the OpenEthereum trace_block and trace_replayBlockTransactions APIs.

This release fixes numerous bugs with this API and now ABI-encodes non-EVM inputs/outputs as if they were explicit EVM calls to handle_filecoin_method for better block explorer compatibility.

However, there are some significant limitations:

  1. The Geth APIs are not implemented, only the OpenEthereum (Erigon, etc.) APIs.
  2. Block rewards are not (yet) included in the trace.
  3. Selfdestruct operations are not included in the trace.
  4. EVM smart contract "create" events always specify 0xfe as the "code" for newly created EVM smart contracts.

Additionally, Filecoin is not Ethereum no matter how much we try to provide API/tooling compatibility. This API attempts to translate Filecoin semantics into Ethereum semantics as accurately as possible, but it's hardly the best source of data unless you need Filecoin to look like an Ethereum compatible chain. If you're trying to build a new integration with Filecoin, please use the native StateCompute method instead.

GetActorEventsRaw and SubscribeActorEventsRaw

FIP-0049 introduced Actor Events that can be emitted by user programmed actors. FIP-0083 introduces new events emitted by the builtin Verified Registry, Miner and Market Actors. These new events for builtin actors are being activated with network version 22 to coincide with Direct Data Onboarding as defined in FIP-0076 which introduces additional flexibility for data onboarding. Sector, Deal and DataCap lifecycles can be tracked with these events, providing visibility and options for programmatic responses to changes in state.

Actor events are available on message receipts, but can now be retrieved from a node using the new GetActorEventsRaw and SubscribeActorEventsRaw methods. These methods allow for querying and subscribing to actor events, respectively. They depend on the Lotus node both collecting events (with Fevm.Events.RealTimeFilterAPI and Fevm.Events.HistoricFilterAPI) and being enabled with the new configuration option Events.EnableActorEventsAPI. Note that a Lotus node can only respond to requests for historic events that it retains in its event store.

Both GetActorEventsRaw and SubscribeActorEventsRaw take a filter parameter which can optionally filter events on:

  • Addresses of the actor(s) emitting the event
  • Specific Fields within the event
  • FromHeight and ToHeight to filter events by block height
  • TipSetKey to restrict events contained within a specific tipset

GetActorEventsRaw provides a one-time query for actor events, while SubscribeActorEventsRaw provides a long-lived connection (via websockets) to the Lotus node, allowing for real-time updates on actor events. The subscription can be cancelled by the client at any time.

A future Lotus release may include GetActorEvents and SubscribeActorEvents methods which will provide a more user-friendly interface to actor events, including deserialization of event data.

Events Configuration Changes

All configuration options previously under Fevm.Events are now in the top-level Events section along with the new Events.EnableActorEventsAPI option mentioned above. If you have non-default options in [Events] under [Fevm] in your configuration file, please move them to the top-level [Events].

While Fevm.Events.* options are deprecated and replaced by Events.*, any existing custom values will be respected if their new form isn't set, but a warning will be printed to standard error upon startup. Support for these deprecated options will be removed in a future Lotus release, so please migrate your configuration promptly.

GetAllClaims and GetAllAlocations

Additionally the methods GetAllAllocations and GetAllClaims has been added to the Lotus API. These methods lists all the available allocations and claims available in the actor state.

Lotus CLI

The filplus commands used for listing allocations and claims have been updated. If no argument is provided to the either command, they will list out all the allocations and claims in the verified registry actor.
The output list columns have been modified to AllocationID and ClaimID instead of ID.

lotus filplus list-allocations --help
NAME:
   lotus filplus list-allocations - List allocations available in verified registry actor or made by a client if specified

USAGE:
   lotus filplus list-allocations [command options] clientAddress

OPTIONS:
   --expired   list only expired allocations (default: false)
   --json      output results in json format (default: false)
   --help, -h  show help


lotus filplus list-claims --help     
NAME:
   lotus filplus list-claims - List claims available in verified registry actor or made by provider if specified

USAGE:
   lotus filplus list-claims [command options] providerAddress

OPTIONS:
   --expired   list only expired claims (default: false)
   --help, -h  show help

Dependencies

Others

Commit log for v1.26.2: v1.26.1...v1.26.2

Commit log since v1.25.2: v1.25.2...v1.26.2