Conversation
app/app.go
Outdated
| } | ||
|
|
||
| // Returns a mapping of the access index to the channels | ||
| func getChannelsFromSignalMapping(signalMapping map[acltypes.AccessOperation][]CompletionSignal) (channelsMapping [][](chan interface{})){ |
There was a problem hiding this comment.
PR to update Context object sei-protocol/sei-cosmos#27
app/graph.go
Outdated
| FromNodeID: edge.FromNodeID, | ||
| ToNodeID: edge.ToNodeID, | ||
| AccessOperation: *edge.AccessOperation, | ||
| Channel: make(chan interface{}), |
There was a problem hiding this comment.
Thought it would make sense to store the channel here as it's set for both to and from nodes
app/graph.go
Outdated
| FromNodeID: edge.FromNodeID, | ||
| ToNodeID: edge.ToNodeID, | ||
| AccessOperation: *edge.AccessOperation, | ||
| Channel: make(chan interface{}), |
|
|
||
| // Gather Results | ||
| txResults := []*abci.ExecTxResult{} | ||
| for result := range resultChan { |
There was a problem hiding this comment.
this isnt necessarily going to be in the same order as txs that were provided, right? is that safe, or should we also include the tx index with the results so we can sort into the tx index order prior to returning Tx results?
There was a problem hiding this comment.
Yes, this will be in the order that it's completed in - let me double check if the ordering matters
| // Store the Channels in the Context Object for each transaction | ||
| ctx.WithTxBlockingChannels(getChannelsFromSignalMapping(txBlockingSignalsMap)) | ||
| ctx.WithTxCompletionChannels(getChannelsFromSignalMapping(txCompletionSignalingMap)) | ||
|
|
There was a problem hiding this comment.
can we add TODOs to block on blocking channels and writing to completion channels to properly block and release resources
There was a problem hiding this comment.
6fb2273 to
696d40c
Compare
0abefd6 to
b5f8a38
Compare
go.mod
Outdated
|
|
||
| replace ( | ||
| github.com/cosmos/cosmos-sdk => github.com/sei-protocol/sei-cosmos v0.1.67 | ||
| github.com/cosmos/cosmos-sdk => ../sei-cosmos // github.com/sei-protocol/sei-cosmos v0.1.20 |
There was a problem hiding this comment.
Will revert this once the sei-cosmos PR merges and I can bump up the version
| return | ||
| } | ||
|
|
||
| type CompletionSignal struct { |
There was a problem hiding this comment.
Moving this up with all the other types
e348010 to
569c6fc
Compare
| Events: deliverTxResp.Events, | ||
| Codespace: deliverTxResp.Codespace, | ||
| }) | ||
| if err == ErrCycleInDAG { |
There was a problem hiding this comment.
would be good to add metrics over how many times we process synchronously vs concurrently
There was a problem hiding this comment.
I'll add a todo here, looks like we don't have the telemetry lib from master. It would be easier to do once we merge this into master
app/app.go
Outdated
|
|
||
| // Waits for all the transactions to complete | ||
| go func() { | ||
| waitGroup.Wait() |
There was a problem hiding this comment.
we're waiting for the WG in a separate goroutine? doesn't that mean it won't block?
There was a problem hiding this comment.
ah shoot, yeah good catch. We'll need another waitGroup.Wait() in the method too
There was a problem hiding this comment.
ah what i meant is do we need this Wait()? From what I can tell, you essentially want to launch a bunch of goroutines in ln 1000 from this main thread, and block when it's finally all done. If that's the case, then the waitGroup.Wati() in ln 1018 is sufficient to achieve this
There was a problem hiding this comment.
Ahh yeah, it's technically not really needed but more so it's just closing the channel. I guess in our case since we don't really need need to communicate that the data is done then we could just remove this
app/app.go
Outdated
| // For each transaction, start goroutine and deliver TX | ||
| for txIndex, txBytes := range txs { | ||
| waitGroup.Add(1) | ||
| app.ProcessTxConcurrent( |
There was a problem hiding this comment.
dont we want this in a separate goroutine?
app/app.go
Outdated
|
|
||
| // Waits for all the transactions to complete | ||
| go func() { | ||
| waitGroup.Wait() |
There was a problem hiding this comment.
ah what i meant is do we need this Wait()? From what I can tell, you essentially want to launch a bunch of goroutines in ln 1000 from this main thread, and block when it's finally all done. If that's the case, then the waitGroup.Wati() in ln 1018 is sufficient to achieve this
e3ff08b to
a192f47
Compare
2dde61f to
da941f6
Compare
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Create Channels from DAG * lint
* Revert "Update Antehandler resource depedencies (#354)" This reverts commit 6c672fb. * Revert "update staking validation identifier (#353)" This reverts commit 8a17c26. * Revert "Update Oracle Parallel Tx Resource Identifiers (#351)" This reverts commit c9229ab. * Revert "Update IdentifierTemplate for TokenFactory (#350)" This reverts commit e81fc93. * Revert "[loadtest] Loadtest fixes" This reverts commit 68bfab5. * Revert "fix unit tests" This reverts commit 49eb809. * Revert "Fix Bank Send Unit Tests (#348)" This reverts commit 9c11837. * Revert "integrate with sei-cosmos changes (#347)" This reverts commit e1b1790. * Revert "Validate Concurrent Messages + Update BankSend (#345)" This reverts commit e4d3176. * Revert "Add concurrency for BurnMsg and MintMsg (#331)" This reverts commit 4a5cff0. * Revert "Ezhu/granularize staking resources (#343)" This reverts commit 90e288c. * Revert "Oracle parallel (#334)" This reverts commit 67c0db3. * Revert "Ezhu/staking dep gen mappings v2 rebase (#338)" This reverts commit 1324183. * Revert "Refactor Loadtest script (#336)" This reverts commit c8b83c7. * Revert "Optimize genesis account creation script (#320)" This reverts commit a0c0306. * Revert "Wasm gov (#316)" This reverts commit c3b421b. * Revert "Lazy Deposit All Module Accounts During EndBlock (#313)" This reverts commit 5ad4d9b. * Revert "Wasm signal (#305)" This reverts commit 206f2bb. * Revert "Add dependencies for ante handlers that read/write accounts (#314)" This reverts commit 5d39f90. * Revert "Add Gasless decorator back and remove CountTxDecorator (#311)" This reverts commit 93fa4bd. * Revert "Branch another cache for all transactions (#309)" This reverts commit c2ce1ed. * Revert "Add msg send dynamic access ops (#303)" This reverts commit 2c5f14e. * Revert "[ante] Add ante dep generator default behavior (#294)" This reverts commit 0ada1db. * Revert "Add more parallel TX metrics (#296)" This reverts commit 567ffd9. * Revert "Add aclmapping options and mapping folder (#287)" This reverts commit c9522a5. * Revert "Bump sei-cosmos and sei-tendermint for 2.0.0beta (#293)" This reverts commit d9254b6. * Revert "Cherry-pick loadtesting changes and make a fix for parallel tx (#288)" This reverts commit 255fec9. * Revert "[app] refactored graph into acl module (#286)" This reverts commit 9b2d204. * Revert "Add gov proposal handler for acl (#277)" This reverts commit bbb5a50. * Revert "Fix for DAG building switch cases (#282)" This reverts commit 9eebc8e. * Revert "Optimize signals (#280)" This reverts commit fd9a3b6. * Revert "[app] Add behavior to process blocks with gov txs sync (#276)" This reverts commit ea2a906. * Revert "[graph] Move metric to dag builder helper (#275)" This reverts commit c530c02. * Revert "Fixes for parallel TX and metrics (#272)" This reverts commit 30207e7. * Revert "[graph] Add resource hierarchy to dependency graph (#268)" This reverts commit da7c070. * Revert "Revert "Parallel TX Synchrnous Testing"" This reverts commit f343740. * Revert "Parallel TX Synchrnous Testing" This reverts commit b95b44c. * Revert "Dag optimization (#263)" This reverts commit ac3fdbf. * Revert "Create Channels from DAG (#261)" This reverts commit 35b9cf4. * Revert "[accesscontrol] Add dependency dag (#258)" This reverts commit 6721845. * Revert "Register accesscontrol module (#257)" This reverts commit 3e34d34.
## Describe your changes and provide context This replaces the original Upsert in the deferred send from module to account with one that doesn't perform a subtraction against the deferred sends. The callsites were also modified to ensure this deferred withdrawal would ONLY be called IFF we've already tried subtracting from the deferred sends with insuffucient balance AND there is sufficient bank balance. ## Testing performed to validate your change Tests in sei-chain verifying this behavior for known problematic scenarios.
## Describe your changes and provide context This replaces the original Upsert in the deferred send from module to account with one that doesn't perform a subtraction against the deferred sends. The callsites were also modified to ensure this deferred withdrawal would ONLY be called IFF we've already tried subtracting from the deferred sends with insuffucient balance AND there is sufficient bank balance. ## Testing performed to validate your change Tests in sei-chain verifying this behavior for known problematic scenarios.
Creates channels based on the DAG
Pointing the PR to #258
Tested by running: ./scripts/old_initialize_local.sh