mongocxx::client::Database::Collection::insert_many() throws exception for 20 to 30 seconds at start up

Question

I have two docker containers that run and communicate with each other. One is a mongo database and one runs c++ mongocxx client in Linux.

In my c++, I have this code which gets called very shortly after startup:

           //queues is a map of database/collection to documents to add or update
           void update_mongo(std::map<std::pair<std::string, std::string>,
                                        std::vector<bsoncxx::document::value>>
                                   queues) 
           {
                mongocxx::instance instance;

                char const* server_ip = std::getenv("MONGO_IP");
                if (!server_ip)
                {
                    server_ip = "localhost";
                }
                mongocxx::uri server_uri(
                    fmt::format("mongodb://{}:{}@{}/?directConnection=true",
                                USERNAME, PASSWORD, server_ip));
                mongocxx::client client(server_uri);

                while (true)
                {
                    for (auto const& [database_collection, documents] : queues)
                    {
                        auto const& [database, collection] =
                            database_collection;

                        std::vector<bsoncxx::document::view> inserts;
                        std::vector<UpdateInfo> updates;

                        for (const auto& document : documents)
                        {
                            // update documents if they exist else add to inserts vector
                        }

                        if (!inserts.empty())
                        {
                            try
                            {
                                client.database(database)
                                    .collection(collection)
                                    .insert_many(inserts);
                            }
                            catch (std::exception const& e)
                            {
                                std::cerr <<
                                    "Error inserting document into Mongo: " <<
                                    std::string(e.what());
                            }
                        }
                   }
               }
           }

However, when I startup for the first 20 to 30 seconds I get very fast repeated logs that look like this:

Error inserting document into Mongo: No suitable servers found (`serverSelectionTryOnce` set): [connection refused calling ismaster on 'mongo:27017']: generic server error
Error inserting document into Mongo: No servers yet eligible for rescan: generic server error
Error inserting document into Mongo: No servers yet eligible for rescan: generic server error
Error inserting document into Mongo: No servers yet eligible for rescan: generic server error
Error inserting document into Mongo: No servers yet eligible for rescan: generic server error
etc...

Eventually, these errors go away.

On the mongo side, my supervisor.conf looks like this:

[program:mongod]
command=/usr/local/bin/docker-entrypoint.sh mongod --quiet --logpath %(ENV_MONGO_LOG)s --logRotate reopen --logappend --bind_ip_all --keyFile %(ENV_KEY_FILE)s
autostart=true
autorestart=true

And in the mongo container, I start it with this:

# Run logrotate on startup
logrotate -f /etc/logrotate.d/mongo

cp -f /etc/supervisor/conf.d/supervisor.base.conf /etc/supervisor/conf.d/supervisord.conf
if [ "$NO_REPLICA" = "1" ]; then
    cat /etc/supervisor/conf.d/supervisor.no.replica.conf >> /etc/supervisor/conf.d/supervisord.conf
else
    cat /etc/supervisor/conf.d/supervisor.replica.conf >> /etc/supervisor/conf.d/supervisord.conf
fi

/usr/bin/supervisord -c /etc/supervisor/conf.d/supervisord.conf

Any idea why it takes so long for the c++ to properly connect to mongo and insert documents? Should I just add a sleep for 30 seconds before actually making the connection?