- Events* logs (stored as a blob) ** variable *** tags array *** message ** plugin lines ** could have debug lines * metrics (stored as a blob) ** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources ** containment_path (corresponds to depth of resource in catalog graph) ** events array (stored as resource_events) *** most variable **** old_value **** new_value **** message *** ever more than 1 event?
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports. Also need to think about unchanged resources. So generate deal with that, or should benchmark mutate reports with that flag on or off?
* logs (stored as a blob) ** variable *** tags array *** message ** plugin lines ** could have debug lines * metrics (stored as a blob) ** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources ** containment_path (corresponds to depth of resource in catalog graph) ** events array (stored as resource_events) *** most variable **** old_value **** new_value **** message *** ever more than 1 event?
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports.
Also need to think about unchanged resources. So Does generate deal with that, or should benchmark mutate reports with that flag on or off?
* logs (stored as a blob) ** variable *** tags array *** message ** plugin lines ** could have debug lines * metrics (stored as a blob) ** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources ** containment_path (corresponds to depth of resource in catalog graph) ** events array (stored as resource_events) *** most variable **** old_value **** new_value **** message
*** ever more than 1 one event? per resource property change, (file content, owner, mode, for instance would be three events), but typically just one
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports.
Also need to think about unchanged resources. Does generate deal with that, or should benchmark mutate reports with that flag on or off?
** variable *** tags array *** message ** plugin lines ** could have debug lines
* metrics (stored as a blob in reports.metrics)
** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources
** containment_path stored as a blob (corresponds to depth of resource in catalog graphreports.resources) ** events array ( but changed resource propertiesstored as in resource_events) *** most variable
**** old_value **** new_value **** message
*** so each resource events array has one event per resource property change, (file content, owner, mode, for instance would be three events), but typically just one **** containment_path (corresponds to depth of resource in catalog graph) **** most variable ***** old_value ***** new_value ***** message
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports.
Also need to think about unchanged resources. Does generate deal with that, or should benchmark mutate reports with that flag on or off?
* logs (stored as a blob in reports.logs) ** variable *** tags array *** message ** plugin lines ** could have debug lines
*** a pe primary report with --debug, for example, adds ~11000 log entries and 5MB to the report. * metrics (stored as a blob in reports.metrics)
** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources
** stored as a blob (reports.resources) ** but changed resource properties stored in resource_events
*** so each resource events array has one event per resource property change, (file content, owner, mode, for instance would be three events), but typically just one **** containment_path (corresponds to depth of resource in catalog graph) **** most variable ***** old_value ***** new_value ***** message
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports.
Also need to think about unchanged resources. Does generate deal with that, or should benchmark mutate reports with that flag on or off?
* logs (stored as a blob in reports.logs) ** variable *** tags array *** message ** plugin lines ** could have debug lines
*** a pe primary report with --debug, for example, adds ~11000 log entries and 5MB to the report.
* metrics (stored as a blob in reports.metrics) ** elements are simple structure (hash of name, value, category) ** count of metrics somewhat variable based on type? * resources ** stored as a blob (reports.resources) ** but changed resource properties stored in resource_events *** so each resource events array has one event per resource property change, (file content, owner, mode, for instance would be three events), but typically just one **** containment_path (corresponds to depth of resource in catalog graph) **** most variable ***** old_value ***** new_value ***** message
Probably simplest to generate reports from the generated catalogs since that provides us with resources and structure already. Also makes the sample dataset more cohesive.
How many variations to build per catalog? Since a report is a snapshot of change in time, could build several variants. Need to check and see how benchmark varies reports. Benchmark varies timestamps, but not events.
Also need to think about unchanged resources. Does generate deal with that, or should benchmark mutate reports with that flag on or off? Will add a flag to generate for now.