There are lots of errors in the log (fine), lots of connections are attempted, leading to a high memory usage (at least in the mirage stack).
The failure scenarios in the server implementation should be rethought to reduce the resource consumption.
A related issue is that once a zone expired (i.e. primary is not reachable for a long time), there's no path to recover / retry to retrieve that zone (AFAICT).
There are lots of errors in the log (fine), lots of connections are attempted, leading to a high memory usage (at least in the mirage stack).
The failure scenarios in the server implementation should be rethought to reduce the resource consumption.
A related issue is that once a zone expired (i.e. primary is not reachable for a long time), there's no path to recover / retry to retrieve that zone (AFAICT).