slapos_crm: Merge ComputeNode_check* alarms into project based alarm
Refactor the implementation of the alarms: - Search from Project rather them query compute nodes, directly respecting if the project can create tickets - Split ticket creation from messaging introducing *_getReportedErrorDict to calculate the error messages (so we can re-use them) - Merge 2 alarms (for check compute node and check compute software installations) info a single one, so we launch at least half the amount of activities for compute nodes. - Drop the specific alarm and to issue a ticket if the user has an instance on a 'close/forever' compute node - Reimplement SupportRequest_recheckMonitoring Merge SoftwareInstallation_getReportedErrorDict into ComputeNode_getReportedErrorDict: We only check for problems from Compute Node perspective, so we should report only one ticket per time. SupportRequest_recheckMonitoring can guarantee that the compute node don't have other problems while closing the ticket. So, this reduces the scope of checks, for example, if the computer isn't connecting, there is no point in check all software installations inside. Another example is, if multiple software releases are failing we dont need to report each one, since the administrator is informed and create multiple tickets for the same compute node will only spam the administrator.
Showing
Please register or sign in to comment