PostgresError: DbHandler exited

[jumski]

This error pops up from time to time:

[Error] Timeout-triggered archive failed: PostgresError: DbHandler exited
    at ErrorResponse (https://deno.land/x/[email protected]/src/connection.js:791:26)
    at handle (https://deno.land/x/[email protected]/src/connection.js:477:6)
    at data (https://deno.land/x/[email protected]/src/connection.js:318:9)
    at https://deno.land/x/[email protected]/polyfills.js:138:30
    at Array.forEach (<anonymous>)
    at call (https://deno.land/x/[email protected]/polyfills.js:138:16)
    at success (https://deno.land/x/[email protected]/polyfills.js:98:9)
    at eventLoopTick (ext:core/01_core.js:168:7) {
  name: "PostgresError",
  severity_local: "FATAL",
  severity: "FATAL",
  code: "XX000"
}

This issue appears to be related to abruptly terminated SQL connections in Session Mode. When this error occurs, it prevents the system from spawning new instances.

I verified the number of connections and I confirm that the worker does not properly release them when terminated.

The solution would be to set a low idle timeout in supavisor or give workers more time to terminate SQL connections properly

[jumski]

I managed to solve it in #12 by using transaction mode.
It is still possible to exhaust the connection pool (see docs), but it should not happen for non-extreme usage.