Abstract: A technology is described for invoking a command over a set of computing instances. An example method may include receiving a request to invoke a command over a set of computing instances managed within a service provider environment. In response, the computing instances included in the set of computing instances having an attribute may be identified and the command may be sent to the computing instances according to a send rate parameter specifying a rate at which the command is sent to a portion of the computing instances. Execution status indications may be received from the computing instances, wherein a number of errors indicated by the computing instances that exceeds an error threshold terminates execution of the command.

Abstract: Mechanisms are presented by which a customer of a resource provider can determine information about the portions of shared resources that are allocated to the customer. The information can include, for example, the software, operating systems software versions, software updates, software patches, and other such packages installed and/or executing on the various resources. In at least some embodiments, a customer can select one of a set of predefined actions that enables the customer to query the allocations, or resource instances, to obtain the information, such as through a series of inventory reports stored to a central location. In at least some embodiments the information can be obtained using a set of gatherers each configured to obtain a specific type of information.

Abstract: A method for communicating messages between a service and a virtual machine instance may include receiving a message from a service. The message includes a target instance identification of at least one virtual machine instance (VMI) to receive the message. A notification of the message and the target instance identification is stored in a message queue for each of a plurality of host servers associated with a plurality of VMIs. The message is stored in a message store without storing at least a portion of the message in the message queue. A request for undelivered messages is received from a VMI associated with a customer account. In response to the request, it is determined that a local cache contains an indication that the message is undelivered, where the local cache is within at least one of the plurality of host server computers associated with the VMI sending the request.