upgrade

Documentation

Overview

Package upgrade provides a Cosmos SDK module that can be used for smoothly upgrading a live Cosmos chain to a new software version. It accomplishes this by providing a BeginBlocker hook that prevents the blockchain state machine from proceeding once a pre-defined upgrade block time or height has been reached. The module does not prescribe anything regarding how governance decides to do an upgrade, but just the mechanism for coordinating the upgrade safely. Without software support for upgrades, upgrading a live chain is risky because all of the validators need to pause their state machines at exactly the same point in the process. If this is not done correctly, there can be state inconsistencies which are hard to recover from.

General Workflow

Let's assume we are running v0.38.0 of our software in our testnet and want to upgrade to v0.40.0. How would this look in practice? First of all, we want to finalize the v0.40.0 release candidate and there install a specially named upgrade handler (eg. "testnet-v2" or even "v0.40.0"). An upgrade handler should be defined in a new version of the software to define what migrations to run to migrate from the older version of the software. Naturally, this is app-specific rather than module specific, and must be defined in `app.go`, even if it imports logic from various modules to perform the actions. You can register them with `upgradeKeeper.SetUpgradeHandler` during the app initialization (before starting the abci server), and they serve not only to perform a migration, but also to identify if this is the old or new version (eg. presence of a handler registered for the named upgrade).

Once the release candidate along with an appropriate upgrade handler is frozen, we can have a governance vote to approve this upgrade at some future block time or block height (e.g. 200000). This is known as an upgrade.Plan. The v0.38.0 code will not know of this handler, but will continue to run until block 200000, when the plan kicks in at BeginBlock. It will check for existence of the handler, and finding it missing, know that it is running the obsolete software, and gracefully exit.

Generally the application binary will restart on exit, but then will execute this BeginBlocker again and exit, causing a restart loop. Either the operator can manually install the new software, or you can make use of an external watcher daemon to possibly download and then switch binaries, also potentially doing a backup. An example of such a daemon is https://github.com/regen-network/cosmosd/ described below under "Automation".

When the binary restarts with the upgraded version (here v0.40.0), it will detect we have registered the "testnet-v2" upgrade handler in the code, and realize it is the new version. It then will run the upgrade handler and *migrate the database in-place*. Once finished, it marks the upgrade as done, and continues processing the rest of the block as normal. Once 2/3 of the voting power has upgraded, the blockchain will immediately resume the consensus mechanism. If the majority of operators add a custom `do-upgrade` script, this should be a matter of minutes and not even require them to be awake at that time.

Integrating With An App

Setup an upgrade Keeper for the app and then define a BeginBlocker that calls the upgrade keeper's BeginBlocker method:

func (app *myApp) BeginBlocker(ctx sdk.Context, req abci.RequestBeginBlock) abci.ResponseBeginBlock {
	app.upgradeKeeper.BeginBlocker(ctx, req)
	return abci.ResponseBeginBlock{}
}

The app must then integrate the upgrade keeper with its governance module as appropriate. The governance module should call ScheduleUpgrade to schedule an upgrade and ClearUpgradePlan to cancel a pending upgrade.

Performing Upgrades

Upgrades can be scheduled at either a predefined block height or time. Once this block height or time is reached, the existing software will cease to process ABCI messages and a new version with code that handles the upgrade must be deployed. All upgrades are coordinated by a unique upgrade name that cannot be reused on the same blockchain. In order for the upgrade module to know that the upgrade has been safely applied, a handler with the name of the upgrade must be installed. Here is an example handler for an upgrade named "my-fancy-upgrade":

app.upgradeKeeper.SetUpgradeHandler("my-fancy-upgrade", func(ctx sdk.Context, plan upgrade.Plan) {
	// Perform any migrations of the state store needed for this upgrade
})

This upgrade handler performs the dual function of alerting the upgrade module that the named upgrade has been applied, as well as providing the opportunity for the upgraded software to perform any necessary state migrations. Both the halt (with the old binary) and applying the migration (with the new binary) are enforced in the state machine. Actually switching the binaries is an ops task and not handled inside the sdk / abci app.

Halt Behavior

Before halting the ABCI state machine in the BeginBlocker method, the upgrade module will log an error that looks like:

UPGRADE "<Name>" NEEDED at height <NNNN>: <Info>

where Name are Info are the values of the respective fields on the upgrade Plan.

To perform the actual halt of the blockchain, the upgrade keeper simply panics which prevents the ABCI state machine from proceeding but doesn't actually exit the process. Exiting the process can cause issues for other nodes that start to lose connectivity with the exiting nodes, thus this module prefers to just halt but not exit.

Automation and Plan.Info

We have deprecated calling out to scripts, instead with propose https://github.com/regen-network/cosmosd as a model for a watcher daemon that can launch gaiad as a subprocess and then read the upgrade log message to swap binaries as needed. You can pass in information into Plan.Info according to the format specified here https://github.com/regen-network/cosmosd/blob/master/README.md#auto-download . This will allow a properly configured cosmsod daemon to auto-download new binaries and auto-upgrade. As noted there, this is intended more for full nodes than validators.

Cancelling Upgrades

There are two ways to cancel a planned upgrade - with on-chain governance or off-chain social consensus. For the first one, there is a CancelSoftwareUpgrade proposal type, which can be voted on and will remove the scheduled upgrade plan. Of course this requires that the upgrade was known to be a bad idea well before the upgrade itself, to allow time for a vote. If you want to allow such a possibility, you should set the upgrade height to be 2 * (votingperiod + depositperiod) + (safety delta) from the beginning of the first upgrade proposal. Safety delta is the time available from the success of an upgrade proposal and the realization it was a bad idea (due to external testing). You can also start a CancelSoftwareUpgrade proposal while the original SoftwareUpgrade proposal is still being voted upon, as long as the voting period ends after the SoftwareUpgrade proposal.

However, let's assume that we don't realize the upgrade has a bug until shortly before it will occur (or while we try it out - hitting some panic in the migration). It would seem the blockchain is stuck, but we need to allow an escape for social consensus to overrule the planned upgrade. To do so, we are adding a --unsafe-skip-upgrade flag to the start command, which will cause the node to mark the upgrade as done upon hiting the planned upgrade height, without halting and without actually performing a migration. If over two-thirds run their nodes with this flag on the old binary, it will allow the chain to continue through the upgrade with a manual override. (This must be well-documented for anyone syncing from genesis later on).

(Skip-upgrade flag is in a WIP PR - will update this text when merged ^^)

Index

• Constants
• Variables
• func BeginBlocker(k Keeper, ctx sdk.Context, _ abci.RequestBeginBlock)
• func NewSoftwareUpgradeProposalHandler(k Keeper) govtypes.Handler
• type AppModule
  • func NewAppModule(keeper Keeper) AppModule
  • func (am AppModule) BeginBlock(ctx sdk.Context, req abci.RequestBeginBlock)
  • func (am AppModule) EndBlock(ctx sdk.Context, _ abci.RequestEndBlock) []abci.ValidatorUpdate
  • func (am AppModule) ExportGenesis(ctx sdk.Context) json.RawMessage
  • func (am AppModule) InitGenesis(ctx sdk.Context, data json.RawMessage) []abci.ValidatorUpdate
  • func (am AppModule) NewHandler() sdk.Handler
  • func (am AppModule) NewQuerierHandler() sdk.Querier
  • func (AppModule) QuerierRoute() string
  • func (AppModule) RegisterInvariants(_ sdk.InvariantRegistry)
  • func (AppModule) Route() string
• type AppModuleBasic
  • func (AppModuleBasic) DefaultGenesis() json.RawMessage
  • func (AppModuleBasic) GetQueryCmd(cdc *codec.Codec) *cobra.Command
  • func (AppModuleBasic) GetTxCmd(cdc *codec.Codec) *cobra.Command
  • func (AppModuleBasic) Name() string
  • func (AppModuleBasic) RegisterCodec(cdc *codec.Codec)
  • func (AppModuleBasic) RegisterRESTRoutes(ctx context.CLIContext, r *mux.Router)
  • func (AppModuleBasic) ValidateGenesis(bz json.RawMessage) error
• type CancelSoftwareUpgradeProposal
• type Keeper
• type Plan
• type QueryAppliedParams
• type SoftwareUpgradeProposal
• type UpgradeHandler

Constants

const (
	ModuleName                        = types.ModuleName
	RouterKey                         = types.RouterKey
	StoreKey                          = types.StoreKey
	QuerierKey                        = types.QuerierKey
	PlanByte                          = types.PlanByte
	DoneByte                          = types.DoneByte
	ProposalTypeSoftwareUpgrade       = types.ProposalTypeSoftwareUpgrade
	ProposalTypeCancelSoftwareUpgrade = types.ProposalTypeCancelSoftwareUpgrade
	QueryCurrent                      = types.QueryCurrent
	QueryApplied                      = types.QueryApplied
)

Variables

var (
	// functions aliases
	RegisterCodec                    = types.RegisterCodec
	PlanKey                          = types.PlanKey
	NewSoftwareUpgradeProposal       = types.NewSoftwareUpgradeProposal
	NewCancelSoftwareUpgradeProposal = types.NewCancelSoftwareUpgradeProposal
	NewQueryAppliedParams            = types.NewQueryAppliedParams
	NewKeeper                        = keeper.NewKeeper
	NewQuerier                       = keeper.NewQuerier
)

Functions

func BeginBlocker

func BeginBlocker(k Keeper, ctx sdk.Context, _ abci.RequestBeginBlock)

BeginBlock will check if there is a scheduled plan and if it is ready to be executed. If the current height is in the provided set of heights to skip, it will skip and clear the upgrade plan. If it is ready, it will execute it if the handler is installed, and panic/abort otherwise. If the plan is not ready, it will ensure the handler is not registered too early (and abort otherwise).

The purpose is to ensure the binary is switched EXACTLY at the desired block, and to allow a migration to be executed if needed upon this switch (migration defined in the new binary) skipUpgradeHeightArray is a set of block heights for which the upgrade must be skipped

func NewSoftwareUpgradeProposalHandler

func NewSoftwareUpgradeProposalHandler(k Keeper) govtypes.Handler

NewSoftwareUpgradeProposalHandler creates a governance handler to manage new proposal types. It enables SoftwareUpgradeProposal to propose an Upgrade, and CancelSoftwareUpgradeProposal to abort a previously voted upgrade.

Types

type AppModule

type AppModule struct {
	AppModuleBasic
	// contains filtered or unexported fields
}

AppModule implements the sdk.AppModule interface

func NewAppModule

func NewAppModule(keeper Keeper) AppModule

NewAppModule creates a new AppModule object

func (AppModule) BeginBlock

func (am AppModule) BeginBlock(ctx sdk.Context, req abci.RequestBeginBlock)

BeginBlock calls the upgrade module hooks

CONTRACT: this is registered in BeginBlocker *before* all other modules' BeginBlock functions

func (AppModule) EndBlock

func (am AppModule) EndBlock(ctx sdk.Context, _ abci.RequestEndBlock) []abci.ValidatorUpdate

EndBlock does nothing

func (AppModule) ExportGenesis

func (am AppModule) ExportGenesis(ctx sdk.Context) json.RawMessage

ExportGenesis is always empty, as InitGenesis does nothing either

func (AppModule) InitGenesis

func (am AppModule) InitGenesis(ctx sdk.Context, data json.RawMessage) []abci.ValidatorUpdate

InitGenesis is ignored, no sense in serializing future upgrades

func (AppModule) NewHandler

func (am AppModule) NewHandler() sdk.Handler

NewHandler is empty, as we do not handle Messages (just proposals)

func (AppModule) NewQuerierHandler

func (am AppModule) NewQuerierHandler() sdk.Querier

NewQuerierHandler registers a query handler to respond to the module-specific queries

func (AppModule) QuerierRoute

func (AppModule) QuerierRoute() string

QuerierRoute returns the route we respond to for abci queries

func (AppModule) RegisterInvariants

func (AppModule) RegisterInvariants(_ sdk.InvariantRegistry)

RegisterInvariants does nothing, there are no invariants to enforce

func (AppModule) Route

func (AppModule) Route() string

Route is empty, as we do not handle Messages (just proposals)

type AppModuleBasic

type AppModuleBasic struct{}

AppModuleBasic implements the sdk.AppModuleBasic interface

func (AppModuleBasic) DefaultGenesis

func (AppModuleBasic) DefaultGenesis() json.RawMessage

DefaultGenesis is an empty object

func (AppModuleBasic) GetQueryCmd

func (AppModuleBasic) GetQueryCmd(cdc *codec.Codec) *cobra.Command

GetQueryCmd returns the cli query commands for this module

func (AppModuleBasic) GetTxCmd

func (AppModuleBasic) GetTxCmd(cdc *codec.Codec) *cobra.Command

GetTxCmd returns the transaction commands for this module

func (AppModuleBasic) Name

func (AppModuleBasic) Name() string

Name returns the ModuleName

func (AppModuleBasic) RegisterCodec

func (AppModuleBasic) RegisterCodec(cdc *codec.Codec)

RegisterCodec registers the upgrade types on the amino codec

func (AppModuleBasic) RegisterRESTRoutes

func (AppModuleBasic) RegisterRESTRoutes(ctx context.CLIContext, r *mux.Router)

RegisterRESTRoutes registers all REST query handlers

func (AppModuleBasic) ValidateGenesis

func (AppModuleBasic) ValidateGenesis(bz json.RawMessage) error

ValidateGenesis is always successful, as we ignore the value

type CancelSoftwareUpgradeProposal

type CancelSoftwareUpgradeProposal = types.CancelSoftwareUpgradeProposal

type Keeper

type Keeper = keeper.Keeper

type Plan

type Plan = types.Plan

type QueryAppliedParams

type QueryAppliedParams = types.QueryAppliedParams

type SoftwareUpgradeProposal

type SoftwareUpgradeProposal = types.SoftwareUpgradeProposal

type UpgradeHandler

type UpgradeHandler = types.UpgradeHandler //nolint:golint