main · last commit
1 month ago ·
2npkec0r
tarantool-etcd-jaa Lease expiry not rate-limited or randomized (thundering-herd revokes)
Lined Up
bd update tarantool-etcd-jaa --claim
bd close tarantool-etcd-jaa
| Created by | Eugene Blikh |
| Owner | bigbes@gmail.com |
| Created | 2026-07-17T13:21:27Z |
| Updated | 2026-07-17T13:21:27Z |
etcd caps lease revocation at ~1000/s (leaseRevokeRate) and randomizes expiry across the TTL window on Promote(). app/etcd/lease.lua:345-423 expiry_loop revokes the ENTIRE expired set every 500ms tick, uncapped and unjittered; rebuild_deadlines assigns deterministic now+ttl. Impact: many same-TTL leases (the common case — a fleet of clients using one TTL) expire as a single burst of DELETE events, spiking watch delivery and WAL. Monotonic clock and full-TTL-on-promotion ARE correct (lease.lua:334-340, wired at roles/etcd.lua:359-367). Interacts with the box.watch edge-guard already documented for rebuild_deadlines — do not let jitter re-refresh live leases.
Revokes capped per tick; rebuild_deadlines jitters across the TTL window without extending live leases; burst test shows spread delivery.
No outgoing dependencies.
Nothing depends on this issue.
| id | tarantool-etcd-jaa |
| content_hash | 644f936e370ee377fab7f6a0569f452f02d3b366492ccd8a76a5105a25d0c051 |
| title | Lease expiry not rate-limited or randomized (thundering-herd revokes) |
| description | etcd caps lease revocation at ~1000/s (leaseRevokeRate) and randomizes expiry across the TTL window on Promote(). app/etcd/lease.lua:345-423 expiry_loop revokes the ENTIRE expired set every 500ms tick, uncapped and unjittered; rebuild_deadlines assigns deterministic now+ttl. Impact: many same-TTL leases (the common case — a fleet of clients using one TTL) expire as a single burst of DELETE events, spiking watch delivery and WAL. Monotonic clock and full-TTL-on-promotion ARE correct (lease.lua:334-340, wired at roles/etcd.lua:359-367). Interacts with the box.watch edge-guard already documented for rebuild_deadlines — do not let jitter re-refresh live leases. |
| design | |
| acceptance_criteria | Revokes capped per tick; rebuild_deadlines jitters across the TTL window without extending live leases; burst test shows spread delivery. |
| notes | |
| status | open |
| priority | 3 |
| issue_type | bug |
| assignee | NULL |
| estimated_minutes | NULL |
| created_at | 2026-07-17T13:21:27Z |
| created_by | Eugene Blikh |
| owner | bigbes@gmail.com |
| updated_at | 2026-07-17T13:21:27Z |
| closed_at | NULL |
| closed_by_session | |
| external_ref | NULL |
| spec_id | |
| compaction_level | 0 |
| compacted_at | NULL |
| compacted_at_commit | NULL |
| original_size | NULL |
| sender | |
| ephemeral | 0 |
| wisp_type | |
| pinned | 0 |
| is_template | 0 |
| mol_type | |
| work_type | |
| source_system | |
| metadata | <binary> |
| source_repo | |
| close_reason | |
| event_kind | |
| actor | |
| target | |
| payload | |
| await_type | |
| await_id | |
| timeout_ns | 0 |
| waiters | |
| hook_bead | |
| role_bead | |
| agent_state | |
| last_activity | NULL |
| role_type | |
| rig | |
| due_at | NULL |
| defer_until | NULL |
| no_history | 0 |
| started_at | NULL |
| is_blocked | 0 |
| issue_id | tarantool-etcd-jaa |
| label | etcd-changelog-sweep |
| id | 019f703d-350a-77a1-9ab2-76f6595c5182 |
| issue_id | tarantool-etcd-jaa |
| event_type | created |
| actor | Eugene Blikh |
| old_value | |
| new_value | |
| comment | NULL |
| created_at | 2026-07-17T16:21:27Z |
| id | 019f703d-350b-7305-aa40-f282ff45cfff |
| issue_id | tarantool-etcd-jaa |
| event_type | label_added |
| actor | Eugene Blikh |
| old_value | NULL |
| new_value | NULL |
| comment | Added label: etcd-changelog-sweep |
| created_at | 2026-07-17T16:21:27Z |
No comments.