Exporter web page hang when one or more databases is down

Details:

• Exporter version 0.8.0
• Postgres ver 8 - 10

Expected:
Data from other databases, even that one is down.

Observation:
No data in prometheus when monitoring one or multiple database, and a database go down/offline

Reproduce:

1. start exporter against fake database (simulate that database is down. Same result when testing on real database).

docker run --net=host -e DATA_SOURCE_NAME="postgresql://db.xxx.com/dbname?sslmode=disable" wrouesnel/postgres_exporter

2. curl or browse http://localhost:9187/metrics and refresh page e.g. 5 - 10 times.

Result:
Browser hangs = no response for 6 sec x amount of refresh/request. Only reply/data when all requests are processed. So prometheus scape will fail. Use default timeout in prometheus (10 sec).

The exporter appears to buffer requests and only comes back with responses when all requests are processed / attempted. Time between each request is approx. 6 sec. (sourcecode: 3 retries) if it cannot access the database(s).

No source file or test is named. Start at the /metrics HTTP scrape entry point and trace the three database retries described in the issue; done means an unavailable database no longer blocks metrics from other databases and scrapes receive a response within the configured timeout.