[bug] AO wipe at API-server re-connect (untrapped)
Closed this issue · 8 comments
Caught on tape, twice and slightly different.
(use full-screen view of the stream video to target the location more easily)
timestamp 10:11:35 onwards
https://www.screencast.com/t/XxSPqqNh
timestamp 4:48:48 shows a some sort of re-sync which results in a kill of the AO.
Different from the screencast above, this case has the HF-reconnect at 04:47:10 indicated with GUI lables and the AO survives that. Only a bit later a separate API-server reload wipes them out in complete silence.
https://www.screencast.com/t/QkLzHeMAUIe
fix:
not sure, the reason for the reconnect not understood. If due arbitrary refresh from BFX server, why cancel the AOs and not keep them?
improvement:
- log the cancellation of orders anytime this happens
suggestion :
- display the modal 'connectivity issue: restart'
- restart the app
- display the 'AO reload' modal.
Edited the report above for clarity.
Stale issue message
Behaviour unchanged so far.
Confirmed on 31.2
Possibly, it's a long-standing issue. For the moment the 'meta-reload' events are still a thing. No definite pattern was found, although mostly they happen in bursts and affect mulitple geo-dis-located instances around the same time.
Given the new changes in 3.31.4 could I suggest we close this one and merge it with #813?
Agreed. With the recent build, the cancellations are logged better and trapped, no longer it is just a silent disappearance.