MastoAPI WS skips notifications/posts when instance under load
Environment
- Installation type (OTP or From Source): from sauce
- Pleroma version (could be found in the "Version" tab of settings in Pleroma-FE): 2.1.50-3451-g1f3c8cf7-shigusegubu https://git.shigusegubu.club/Shigusegubu.club/pleroma/commit/1f3c8cf707575a473981a20146ce98578275ddbb
- Elixir version (
elixir -v
for from source installations, N/A for OTP):
Erlang/OTP 22 [erts-10.7] [source] [64-bit] [smp:4:4] [ds:4:4:10] [async-threads:1]
Elixir 1.10.2 (compiled with Erlang/OTP 21)
- Operating system: debian linux
- PostgreSQL version (
psql -V
):psql (PostgreSQL) 11.5 (Debian 11.5-1+deb10u1)
Bug description
When instance under moderate-to-heavy loads, i.e. running the remove_embedded_objects --vacuum
MastoAPI WS skips some notifications/posts. I'm not sure if it actually skips or WS silently drops and reconnects, probably former because PleromaFE does a terrible job at reconnecting the WS and some notifications did get through.
From user perspective it looks like there are no notifications and no new posts for a while, then a stray notif and post might come in, but it was most noticeable in notifications, where i would see more notifications on husky (https) versus pleromafe in ws mode, refreshing the page obviously also helps.