Closed bodqhrohro closed 7 months ago
Confirmed this is still present in 0.0.9-alpha.1
It looks like this is rooted in the default behavior of the Sled datastore used by the matrix-sdk crate if you write to it too often. Maybe cache more in memory and flush to disk only periodically?
Can't reproduce it on latest dev build.
Can confirm, the sled
worker threads no longer show up in iotop
on iamb 0.0.9-alpha.1 (999399a) and I'm not seeing elevated disk I/O.
Still seeing this on a2a708f1ae1dc8d08620f1b302b2cafb159cb40a in iotop-c
@ErrorNoInternet I updated the SDK in #200 , which now uses SQLite instead of sled. Testing on my laptop at ec81b72, I'm seeing long periods of inactivity, with the occasional spike up to ~70 K/s. Can you check if a more recent build improves things for you?
Ah I had to compile it manually with cargo and that worked. The Nix flake (nix run github:ulyssa/iamb) was out of date somehow. Seeing basically zero disk usage now :)
Possibly in the mmapped
~/.config/iamb/profiles/bq/matrix/matrix-sdk-state/db
, as I see no active write operations instrace
.iotop-c
shows around 1 M/s write all the time.I suppose that's not healthy for my SSD.