-
Notifications
You must be signed in to change notification settings - Fork 707
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Use cached session index to obtain executor params #1190
Conversation
Wouldn't it just make more sense to store the |
That carries a pretty high overhead (as it'd be duplicated for every candidate in the session), and recent sessions' info should be generally available. But yes, it'd be more foolproof. |
* master: (25 commits) fix typos (#1339) Use bandersnatch-vrfs with locked dependencies ref (#1342) Bump bs58 from 0.4.0 to 0.5.0 (#1293) Contracts: `seal0::balance` should return the free balance (#1254) Logs: add extra debug log for negative rep changes (#1205) Added short-benchmarks for cumulus (#1183) [xcm-emulator] Improve hygiene and clean up (#1301) Bump the known_good_semver group with 1 update (#1347) Renames API (#1186) Rename `polkadot-parachain` to `polkadot-parachain-primitives` (#1334) Add README to project root (#1253) Add environmental variable to track decoded instructions (#1320) Fix polkadot-node-core-pvf-prepare-worker build with jemalloc (#1315) Sassafras primitives (#1249) Restructure `dispatch` macro related exports (#1162) backing: move the min votes threshold to the runtime (#1200) Bump zstd from 0.11.2+zstd.1.5.2 to 0.12.4 (#1326) Remove `substrate_test_utils::test` (#1321) remove disable-runtime-api (#1328) [ci] add more jobs for pipeline cancel, cleanup (#1314) ...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Added my suggestion in #1355. Otherwise, I think this is strictly better than status quo, so approving.
bot merge |
@eskimor Unknown command "merge". Refer to help docs and/or source code. |
Why not store the execution parameters on their own in the availability store and let the availability data reference it using a hash? Then we don't need to duplicate in every availability data. |
How would that be better than having them stored on chain? The chain will keep the data far longer than we would in the availability store. The problem (that has been fixed) was that we depended on the state of a particular relay parent to be available. I don't see a problem to rely on long lived data on chain. We already rely on the chain for getting the PVF for example. We also always need access to session information. Hence putting execution environment parameters into availability would not gain us anything in terms of robustness. Also it should really be fine, assuming we have probabilistic finality maintained in all cases. Mooaar tests as pointed out by @rphmeier are in order though. |
Yes. The actual issue is that we were attempting to read out the state of a specific relay parent, whereas our approach here is usually to guarantee that the necessary data will be available in every relay chain state for the foreseeable future. |
* master: (25 commits) Markdown linter (#1309) Update `fmt` file and some authors (#1379) Bump the known_good_semver group with 1 update (#1375) Bump proc-macro-warning from 0.4.1 to 0.4.2 (#1376) feat: add futures api to `TransactionPool` (#1348) Ensure cumulus/bridges is ignored by formatter and run it (#1369) substrate: chain-spec paths corrected in zombienet tests (#1362) contracts: Update to wasmi 0.31 (#1350) [improve docs]: Template pallet (#1280) [xcm-emulator] Unignore cumulus integration tests (#1247) Fix wrong ref counting (#1358) Use cached session index to obtain executor params (#1190) fix typos (#1339) Use bandersnatch-vrfs with locked dependencies ref (#1342) Bump bs58 from 0.4.0 to 0.5.0 (#1293) Contracts: `seal0::balance` should return the free balance (#1254) Logs: add extra debug log for negative rep changes (#1205) Added short-benchmarks for cumulus (#1183) [xcm-emulator] Improve hygiene and clean up (#1301) Bump the known_good_semver group with 1 update (#1347) ...
* Import changes from archieved repo * Revert erroneous changes * Fix more tests * Resolve discussions * Fix MORE tests * approval-voting: launch_approval better interface (#1355) --------- Co-authored-by: Javier Viola <javier@parity.io> Co-authored-by: ordian <noreply@reusable.software> Co-authored-by: ordian <write@reusable.software>
This pull request has been mentioned on Polkadot Forum. There might be relevant details there: https://forum.polkadot.network/t/stalled-parachains-on-kusama-post-mortem/3998/1 |
Closes #579
To validate a candidate, approval voting and dispute coordinator subsystems currently rely on the candidate validation subsystem, which requests executor parameters at a session determined by the candidate's relay parent. This is unreliable, as the state at that relay parent may already be pruned. Still, we want to be able to apply slashes in that case.
This PR makes use of the internal cache maintained by approval voting and dispute distribution subsystems to reliably obtain session index even if the state in question isn't there. Then, the execution parameters are requested and provided to the candidate validation subsystem, which doesn't have to request them on its own anymore.
The positive side effect is that the candidate validation subsystem is returning to its pre-#6161 state; that is, it expects the exhaustive validation data from a requesting subsystem instead of doing its own data collection.
The backing subsystem, also affected by the change, does not maintain any cache of session indexes as it doesn't need it, as the state at that stage is guaranteed to be alive. A session index is always known when a backing job is created. Executor parameters are requested before performing the actual validation using that session index.
A rather large volume of changes is dictated by the fact that every test doing candidate validation needs to be fixed, along with the backing, approval, and dispute tests. The actual logic changes are quite compact.