Currently, we have multiple identical slave tests performed by 4 different agents geographically seperate to ensure accuracy of failed tests. Then a "master test" is performed which is dependent on the failure of all the slave tests (one test per agent).
Using multiple paths (in this case geographically placed agents) ensures that the test is really up/down and not just internet connectivity or a bad route somewhere between HM and the agent.
To make things simple would be the ability to group various agents together and assign a test to an "agent group". So AgentA, AgentB, and AgentC could become Master AgentA. All agent subordinates must report up/down status for Master AgentA to succeed/fail.
Regards
Grouping Agents for Undeniable Test Accuracy
But agent may perform many various not related tests. So may be its better to assign several agents for 1 tests?To make things simple would be the ability to group various agents together and assign a test to an "agent group". So AgentA, AgentB, and AgentC could become Master AgentA. All agent subordinates must report up/down status for Master AgentA to succeed/fail.
Or may be use something like "backup rma" feature but with option to repeat test thru backup rma not only in case of HostMonitor<->RMA connection problems but in any case when test returns Unknown status and in some cases when test returns "bad" status
I think this should depends on test method. E.g. "Bad" status for file related tests and tests like Process, Service, Dominant Process, NT Event Log means RMA was able to reach target server and retrieve information that indicates error so there is no reason to use backup RMA and repeat the test. While "No answer" status for tests like Ping, TCP, UDP can be caused by some network problem between agent and target host, in such case HostMonitor may try to perform test using backup RMA...Using 4 RMA for the same test? Isn't this too much?
On the other hand, we plan to implement another option for distributed monitoring - several HostMonitors should be able to work together. May be we can add this option on HostMonitor's level
Regards
Alex