agent wishlist

The concept of PostgreSQL's management and monitoring on a given server (host/target) depends on information gather derived from the database and correlated with operating system information (CPU, memory, network, disk, & ca ). These statistics are obtained for a Cedrus_Agent, which is a service that remain sleeping for a time determined by a config parameter.

Information obtained in a given interval are a sort of baseline, which can be used to generate statistic data in timelines xor ad-hoc analysis from a server instance.

Several agents on multiple servers can be manipulated using a single Cedrus instance through an RPC's object-oriented implementation. For security reasons these calls should be established through an authenticated proxy inside a TLS (transport layer security).

Agents should support the data collection from distinct protocols, and only the Cedrus native procotol will be supported in our first release, expanding to SNMP in the next releases.

It should support multiple database instances in a particular host and will be a Target specialization (see Target spec) from Cedrus standpoint.
There is also a relationship to PgControl structure (see PgControl) to get records information of WAL stuff.

The service would be responsible for:
* Launch and control signals such as: SIGHUP, SIGINT, SIGQUIT, and SIGTERM SIGABRT;
* Communication to gather Target's historical information;
* Handling file system issues;
* Monitoring system in the Host;
* Monitoring CPU, memory, system load, disk space;
* Notification about Cedrus changes in the host;
* Reading log in host to send information, alerts, errors, checkpoint intervals