From 0ccdd12785fe3be0da5a230a841af357f1ec399f Mon Sep 17 00:00:00 2001 From: Alexey Murz Korepov Date: Fri, 29 May 2020 18:27:07 +0300 Subject: [PATCH] Added short info about Federation section --- ...apse-Performance-Issues-Through-Grafana-Graphs.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/Understanding-Synapse-Performance-Issues-Through-Grafana-Graphs.md b/Understanding-Synapse-Performance-Issues-Through-Grafana-Graphs.md index 78ddefa..dff92d2 100644 --- a/Understanding-Synapse-Performance-Issues-Through-Grafana-Graphs.md +++ b/Understanding-Synapse-Performance-Issues-Through-Grafana-Graphs.md @@ -24,6 +24,18 @@ In the first graph, we can see obvious spikes corresponding to lots of `get_user Still, it's probably worth investigating why we're getting users from the database that often, and whether it's possible to reduce the amount of queries we make by adjusting our cache factor(s). +## Federation + +Federation section charts shows info about incoming and outgoing federation transactions, that have two major types: + +- PDU (Persistent Data Unit) - events, that stored in persistent database: room messages, state events (join/leave), etc. +- EDU (Ephemeral Data Unit) - events, that not need to be stored permanently, such as read receipts, typing notifications. + +Also "Outgoing EDUs by type" chart shows outgoing federation transactions by type: `m.device_list_update`, `m.direct_to_device`, `m.presence`, `m.receipt`, `m.typing`. + +If you see too much `m.presence` transactions with too much CPU load, you can disable `presence` in Synapse config, issue about this problem - here: [Presence is increasingly heavy](https://github.com/matrix-org/synapse/issues/3971) + + ## Caches ![image](https://user-images.githubusercontent.com/1342360/82240572-8b239180-9932-11ea-96ff-6b5f0e57ebe5.png)