brimdata
Reconnect to /events endpoint if the connection is terminated
Noticed this when upgrading the cloud instances and all connections get terminated when the zed service process restarts. The events endpoint gets severed and the client does not attempt to reestablish the connection.
I tried to reproduce this issue to witness the user-facing symptoms of the problem, but was unsuccessful. My specific approach was to study a connection from the app to localhost:9867 so I could easily observe captured traffic in Wireshark. What I thought would repro the problem would be to kill the zed serve process and then restart it, then do a zed create from the CLI, expecting that the pools list in the app would not update to reflect the new pool name. However, in my attempts, it always did update. This was even true when I went back to Brim/Zed combinations back in November of 2021 when this issue was filed.
However, when doing a test permutation connecting to the genuine cloud setup, I instead bumped into what looks like new issue #2545 where my research determined that the /events endpoint access seems to fail completely over a connection to the cloud service (not just after a service restart... always) due to lack of auth. I suspect that once we address that issue we'll find this issue #1972 is not reproducible, but for now I'll just mark this one as "blocked" by #2545 as a reminder to confirm.
