You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
It can be seen in logs and client UIs that syncing frequently times out. This is most likely because of the smaller timeout values set in f26e3ed to address unit tests occasionally failing when the first request gets stuck for some reason. Long polling implies that sync requests may stay with the server for quite some time, and the current 30 seconds are definitely not long enough to support that.
Expected behavior
It's probably best to revert timeouts to their original values and adjust timeouts in unit tests instead. They might run longer but network requests getting stuck are not a normal situation anyway.
The text was updated successfully, but these errors were encountered:
Fun fact: the default timeout for Connection::syncLoop() is also 30 seconds, but it's the server-side time, so the library might be giving up on waiting just before the empty response arrives, due to the network latency. Merely increasing the default timeout to 40s could remove those failures.
SyncJob doesn't only have unlimited retries now, it also backs off in a
different way, to address for its special situation as a de-factor keep-
alive signal. Fixes#837.
SyncJob doesn't only have unlimited retries now, it also backs off in a
different way, to address for its special situation as a de-factor keep-
alive signal. Fixes#837.
Describe the bug
It can be seen in logs and client UIs that syncing frequently times out. This is most likely because of the smaller timeout values set in f26e3ed to address unit tests occasionally failing when the first request gets stuck for some reason. Long polling implies that sync requests may stay with the server for quite some time, and the current 30 seconds are definitely not long enough to support that.
Expected behavior
It's probably best to revert timeouts to their original values and adjust timeouts in unit tests instead. They might run longer but network requests getting stuck are not a normal situation anyway.
The text was updated successfully, but these errors were encountered: