Efficient network fleet monitoring

摘要

Methods and apparatus for efficient monitoring of network fleets are described. A list of network addresses of a set of hosts at which resources are to be monitored from a monitoring server of a provider network may be received at the monitoring server. The monitoring server may initiate establishment of a persistent network connection to a monitoring agent installed at a monitored host. A plurality of health messages from the monitoring agent may be obtained via the connection, including a host status entry for the monitored host and a resource status entry for at least one resource configured at the monitored host. A representation of the health messages may be saved in a repository for analysis.

主权项

1. A system, comprising:
one or more computing devices including a particular monitoring server, wherein the one or more computing devices comprise one or more hardware processors configured to:
receive, at the particular monitoring server of one or more monitoring servers configured to collect health state information of a plurality of network-accessible resources of a provider network, a list of network addresses of a set of monitored hosts of the provider network, wherein said list is generated by an administrative service of the provider network;initiate, from the particular monitoring server, an establishment of a persistent network connection to a monitoring agent installed at a monitored host of the set of monitored hosts;obtain, at the particular monitoring server via the persistent network connection during a time interval, a plurality of health messages from the monitoring agent, wherein the plurality of health messages comprises (a) a host status message associated with the monitored host and (b) a resource status message associated with at least one resource configured at the monitored host;store, at a storage service of the provider network, a representation of the plurality of health messages, wherein the representation comprises (a) a host status entry associated with the host status message and (b) a resource status entry associated with the resource status message;in response to a determination, based at least in part on an analysis of the plurality of health messages, that an unexpected state was encountered at the monitored host, initiate a corrective action corresponding to the unexpected state.