element-hq / element-hq/synapse
federation sender was very slow to catch up after a spam
- Dominant language
- Python
- Stars
- 4.6k
- Forks
- 600
- Avg merge
- 5d 22h
- Merged PRs (30d)
- 51
Description
This issue has been migrated from [#10257](https://github.com/matrix-org/synapse/issues/10257).
---
on the `kde.org` homeserver, somebody sent 4500 events into a room with 415 active servers. The federation sender then spent the next two hours sending out those events one by one.
The question is: why did we only send one event per transaction?
Contributor guide
Research direction
The report concerns the federation sender on the kde.org homeserver, which sent events one per transaction after a 4,500-event spam burst. Start by tracing the federation sender's transaction batching behavior and reproduce the catch-up scenario if possible. Done means the sender batches eligible events and no longer takes hours to process this backlog.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- backend, distributed-systems, networking
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100