-
Notifications
You must be signed in to change notification settings - Fork 745
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
Node approval voting behaviour when disabled (no-show vs explicit vote) #2725
Comments
Only participating once others raised it seems more or less equivalent to disputing immediately in that case. |
Are you saying that
AKA no-show first and participate if someone escalates it is equivalent to
disputing immediately but getting ignored as it is in our current disabling spec / implementation? Those two definitely don't seem equivalent to me. Example: In the current spec of issuing an explicit vote the other nodes will for sure ignore this one desperate node as it is disabled and will not participate further. In case it just no-shows there is still hope that more tranches will bring some honest non-disabled dispute statements. Am I understanding correctly? |
I am not following. In both cases the validator would no-show, triggering more tranches. |
68d8650 Bump thiserror from 1.0.50 to 1.0.51 009c989 remove no longer valid check from the ensure_weights_are_correct (#2740) 94c44a7 Added Rococo BH <> Rococo Bulletin bridge (#2724) 5fe0f2f Bump tokio from 1.34.0 to 1.35.0 25f8251 Grafana update stuff (#2733) 06fbe8b Improved `ExportXcm::validate` implementation for BridgeHubs - step 1 (#2727) 390e836 Select header that will be fully refunded in on-demand batch finality relay (#2729) ce701dd separate constants for average and worst case relay headers (#2728) 09215c5 Backport from `polkadot-sdk` + bump (#2725) 6327261 Bump serde from 1.0.192 to 1.0.193 fff9ddd Bump sysinfo from 0.29.10 to 0.29.11 4be99fe Monitoring and alerts for Rococo/Westend (#2710) 67a683a Bump ed25519-dalek from 2.0.0 to 2.1.0 8e0e794 quick and dirty fix for the `wait -p` and older distros (#2712) 3ab6562 Add withdraw reserve assets to zombienet tests (#2711) c2c409b increase init timeouts in zombienet tests (#2706) a8c60b4 fix lane id and bridged chain id (#2705) 9ac0f26 removed bp-asset-hub-kusama and bp-asset-hub-polkadot (#2703) 4916475 Some fixes for zombienet tests (polkadot-staging) (#2704) 6f9a147 zombienet from Wococo to Westend (#2699) 3ba7910 Porting changes from polkadot-sdk to polkadot-staging - before update subtree with removed wococo stuff (#2696) 653448f Remove Woococo related stuff (#2692) 03aaab2 Gitspiegel polkadot staging (#2695) 702a4c1 Drop Rialto <> Millau bridges (#2663) (#2694) 6a63b5f Start version guards for the ED loop (#2678) 896b9a9 typo (#2690) 671d27c Bump serde from 1.0.190 to 1.0.192 991b229 Bump clap from 4.4.7 to 4.4.8 ec267ec Bump env_logger from 0.10.0 to 0.10.1 592e407 Bump tokio from 1.33.0 to 1.34.0 c49ce3d Bump serde_json from 1.0.107 to 1.0.108 04b3319 Update subxt-codegen version (#2674) 03f9804 backport #2139 (#2673) 49245dd removed unused PARACHAINS_FINALITY_PALLET_NAME constant (#2670) 658a3f5 BHR/BHWE spec_version according to the `polkadot-sdk` (#2668) 7666b94 Nit from `polkadot-sdk` (#2665) b5c43bb Adjusted constant because for measuring we used mistakenly rococo constants (#2664) 062449d Add Rococo<>Westend bridge support/relay (#2647) 55eb44e Add basic zombienet test to be used in the future (#2649) (#2660) 93b6b3f Bump clap from 4.4.6 to 4.4.7 4c01ab0 Bump futures from 0.3.28 to 0.3.29 a31a6c0 Bump tempfile from 3.8.0 to 3.8.1 bcdfe83 Bump serde from 1.0.189 to 1.0.190 f7433b0 Port #2648 to polkadot-staging (#2651) 3896738 Bump scale-info from 2.9.0 to 2.10.0 12d62c5 Bump thiserror from 1.0.49 to 1.0.50 1d78aa1 Backport from `polkadot-sdk` with actual master (#2633) ab4de94 Grandpa justifications: Avoid duplicate vote ancestries (#2634) (#2635) 465562a add missing crate descriptions (#2629) 28d3680 Bump fixed-hash 67528c4 Bump serde from 1.0.188 to 1.0.189 d450c47 Bump time from 0.3.29 to 0.3.30 6a19f83 Bump async-trait from 0.1.73 to 0.1.74 a92d213 Millau, Rialto: accept equivocation reports (#2614) (#2617) a61f777 Bump tokio from 1.32.0 to 1.33.0 0052f64 Bump subxt from 0.32.0 to 0.32.1 ccc849d Bump num-traits from 0.2.16 to 0.2.17 22f2752 apply late suggestions for #2600 (#2603) 0320172 actualize check_obsolete_call comment (#2601) 5cbbd25 Reject transactions if bridge pallets are halted (#2600) ca4dfe3 Bump subxt from 0.31.0 to 0.32.0 8bf7b58 Bump clap from 4.4.4 to 4.4.6 88b0b99 Bump thiserror from 1.0.48 to 1.0.49 263833b https://gitlab.parity.io/parity/mirrors/polkadot-sdk/-/jobs/3833103 (#2589) 4f44968 Backport changes from polkadot-sdk (#2588) 7200ed1 fiox overflow when computing priority boost (#2587) e02cbd3 Bump time from 0.3.28 to 0.3.29 a097dd2 Bump clap from 4.4.3 to 4.4.4 801ce88 Merge bulletin chain changes into polkadot staging (#2574) a3803ce Add unit tests for the equivocation detection loop (#2571) 26dfc31 Bump clap from 4.4.2 to 4.4.3 66a8beb Bump serde_json from 1.0.106 to 1.0.107 18c50da Bump trie-db from 0.27.1 to 0.28.0 4c4fa92 Equivocation detection loop: Reorganize block checking logic as state machine (#2555) (#2557) 6bd317a Bump serde_json from 1.0.105 to 1.0.106 a7e6bfd Backport for polkadot-sdk#1446 (#2546) d9f8050 Bump sysinfo from 0.29.9 to 0.29.10 901f44c Bump thiserror from 1.0.47 to 1.0.48 82eeb50 Bump sysinfo from 0.29.8 to 0.29.9 a0c934b Bump strum from 0.24.1 to 0.25.0 1064fbf Bump subxt from 0.28.0 to 0.31.0 e50398d bridges subtree fixes (#2528) 99af075 Markdown linter (#1309) (#2526) 733ff0f `polkadot-staging` branch: Use polkadot-sdk dependencies (#2524) e8a59f1 Fix benchmark with new XCM::V3 `MAX_INSTRUCTIONS_TO_DECODE` (#2514) 62b185d Backport `polkadot-sdk` changes to `polkadot-staging` (#2518) d9658f4 Fix equivocation detection containers startup (#2516) (#2517) d65db28 Backport: building images from locally built binaries (#2513) 5fdbaf4 Start the equivocation detection loop from the complex relayer (#2507) (#2512) 7fbb67d Backport: Implement basic equivocations detection loop (#2375) cb7efe2 Manually update deps in polkadot staging (#2371) d17981f #2351 to polkadot-staging (#2359) git-subtree-dir: bridges git-subtree-split: 68d8650
Not an issue. Approval Checking statement is always implies a valid candidate. If checker finds an invalid candidate it emits a dispute statement instead and it goes directly to the Dispute Coordinator subsystem. So despite issuing a statement from the perspective of the Approval Checking subsystem it will register as a no-show if given enough time. No-showing is the current behaviour and it is the preferred one because it leads to a soft escalation. |
* Backport from `polkadot-sdk` * Bump `polkadot-sdk` * revert * Cargo.lock
When an (honest) node gets disabled it has limited validator privileges. Even if it spots a truly malicious candidate there is not that much it can do since raising a dispute will not escalate into participation from others (unless confirmed / other non-disabled participated already). Should that node issue an explicit vote in the dispute or should it no-show to slightly escalate the issue? Then if this node sees the dispute for this candidate started it will actually participate.
* Above only concerns finding an invalid candidate. In the happy path of a valid candidate node should always definitely issue an explicit vote for validity.
No-showing might be safer and a better approach if the node truly believes something is wrong as it escalates the issue further than casting an invalid vote when disabled.
Using that approach is equivalent safety-wise to a DoS attack on honest approval votes to potentially silence them. We are fine because of the escalation bringing in even more honest validators on average.
The cost we're paying is that disabled validators will have some escalation tools. So they can still affect liveness if they really want to but this will require much greater numbers. Without disabling a single guy could trigger a dispute (absolute escalation) but with that strategy a single actor if he gets disabled can at most... create 1 no-show (miniscule escalation).
TODOs:
Adjust node behaviour when disabledThe text was updated successfully, but these errors were encountered: