Messages stuck or very slow at "Encrypting your message..."

[germain-gg]

Steps to reproduce

1. Send a message

Outcome

What did you expect?

The message to send instantanously

What happened instead?

It's very slow, above the 2mn mark!

Operating system

No response

Application version

No response

How did you install the app?

No response

Homeserver

No response

Will you send logs?

Yes

[t3chguy]

Duplicate of #16043

[germain-gg]

Not a duplicate, this has just happened in a room with 20 people. I would not consider this a "massive" room.

[germain-gg]

Adding X-Regression this was not happening a month ago!

[clokep]

I've also seen this happen in fairly small rooms. Although mine wasn't quite as small (57 and 103 users). Maybe someone has a lot of devices though?

[daniellekirkwood]

I encountered this today morning.

• I was sending one message to Element Web Internal (27 people)
• I also was sending a few messages in a DM to @clokep
• One message to a room with 3 people (pressed Enter at 11:10am > Message still not sent at 11:15)

I was able to send messages instantly on iOS, on web these 2 rooms took around 5 minutes to send - they stayed looking like the screenshot above.

I upgraded my version using the pop up tile in the top left
this morning before sending any messages.

I'm using Nightly: Element Nightly version: 2023011701 | Olm version: 3.2.12

Sent a rageshake

[germain-gg]

I've seen a few ad hoc mention of this issue in multiple internal rooms. Will bump to O-Frequent as it's been experienced many times by many different people

[ara4n]

Maybe someone has a lot of devices though?

This is "only" 1276 devices. I've benchmarked it in Oct v Jan based on rageshakes for the same number of devices (43 slices, so 1075 in total) and it looks like we have a ~50% slowdown from somewhere. It also really feels like it's churning megolm more frequently than I'd expect, and/or not setting them up proactively enough.

[davidegirardi]

I got this in multiple rooms today, not in DMs for some reason.

I checked the console log in one room where I had 3 "stuck" messages and I found three of these:

rageshake.ts:73 Event ~!EVENTIDHERE in room !ROOMIDHERE is live, but it does not have a timeline
e.<computed> @ rageshake.ts:73

See the rageshake for the event information. At the time of writing they have been stuck for about 2 hours.

I could receive other people's messages just fine.

EDIT: restarting Element Desktop just removed the messages without sending them.

[richvdh]

I did a deep dive on one instance of a slow send at https://github.com/matrix-org/element-web-rageshakes/issues/20102; my conclusion is:

• it takes 80 seconds to send keys out in a room with 1288 devices (~60ms per device).
• There might be room to optimise it, but since we're replacing the offending code as part of Adopt the E2EE cryptography module from matrix-rust-sdk: "Element R" #21972 I'm not sure that is a good use of time.
• I see no evidence of this being a regression, though if someone can point me to logs from October or earlier I can compare.

(If messages are actually getting stuck for periods of several minutes, that sounds like a different problem.)

[davidegirardi]

FYI I attached a (different) rageshake to #15476 too as it seems the closest to my situation.

[richvdh]

Duplicate of #15476.

[richvdh]

(If messages are actually getting stuck for periods of several minutes, that sounds like a different problem.)

Have opened #24612 to track this