Soak: a reconnect once stalled over 30 s at 'Återansluter…'

littlefox commented

2026-10-18 11:41:59 +00:00

Owner

Seen once in a local full e2e run on the #97 branch (desktop, soak.spec.ts), passed on retry and in 3 solo reruns. After a deliberate drop, the reader stayed at 'Anslutningen bröts. Återansluter…' (with 'Försök nu' / 'Logga in igen') for more than 30 s; the relay log shows only clean client closes. Possibly a failed first attempt pushing the backoff to 15–30 s (DELAYS = [1,2,4,8,15,30]) under load. To look at: log each reconnect attempt and its error in the events list, and see whether a failed attempt during a busy moment (e.g. the background recount from #96) explains it.

littlefox added the

kind/bug

area/web

labels

2026-10-18 11:41:59 +00:00

littlefox referenced this issue

2026-10-18 11:42:29 +00:00

Settings: a screen of their own that saves itself #110

littlefox referenced this issue

2026-10-18 12:01:13 +00:00

Confusing app settings #97

littlefox referenced this issue from a commit

2026-10-19 08:09:24 +00:00

Reconnect: time out a stuck login, close failed attempts, log why

littlefox referenced this issue from a pull request that will close it,

2026-10-19 08:09:31 +00:00

Reconnect: time out a stuck login, close failed attempts, log why #145

littlefox closed this issue

2026-10-19 08:41:50 +00:00

littlefox commented

2026-10-19 08:41:56 +00:00

Author

Owner

Merged in #145. The stall itself never reproduced, but three things in the reconnect path could cause it or hide it, and all three are fixed:

Once connected, an attempt had no time limit for logging in and starting. A connection that stopped answering there waited forever. It now gets 20 s.
A failed attempt left its connection open, and those piled up against the relay's per-address limit (8). Failed attempts now close it.
Failures weren't recorded. Each one is now in the events list ('Återanslutningen misslyckades: …'), so the next stall can be read off the screen.

Verified: a new reconnect.spec test where the first attempt gets the greeting and then hangs (given up after 20 s, socket closed, reason shown, the next attempt gets through), soak.spec with 16 drops in 60 s, and CI green. Closing; reopen if it shows up again with the events list in view.

Merged in #145. The stall itself never reproduced, but three things in the reconnect path could cause it or hide it, and all three are fixed: - Once connected, an attempt had no time limit for logging in and starting. A connection that stopped answering there waited forever. It now gets 20 s. - A failed attempt left its connection open, and those piled up against the relay's per-address limit (8). Failed attempts now close it. - Failures weren't recorded. Each one is now in the events list ('Återanslutningen misslyckades: …'), so the next stall can be read off the screen. Verified: a new reconnect.spec test where the first attempt gets the greeting and then hangs (given up after 20 s, socket closed, reason shown, the next attempt gets through), soak.spec with 16 drops in 60 s, and CI green. Closing; reopen if it shows up again with the events list in view.

Rows
Columns

Soak: a reconnect once stalled over 30 s at 'Återansluter…' #109