OpenSolaris

Printable Version Enter a New Search
Bug ID 6311428
Synopsis exacct doesn't account for a process's microstate data when it moves between tasks
State 10-Fix Delivered (Fix available in build)
Category:Subcategory kernel:accounting
Keywords accounting | extacct | extended | newtask | rtiq_reviewed | settaskid
Responsible Engineer Robert Harris
Reported Against 5.9
Duplicate Of
Introduced In solaris_8u1
Commit to Fix snv_69
Fixed In snv_69
Release Fixed solaris_nevada(snv_69) , solaris_10u5(s10u5_03) (Bug ID:2138526)
Related Bugs 6455548
Submit Date 16-August-2005
Last Update Date 19-July-2007
Description 
The extended accounting subsystem (exacct) sometimes reports unrealistically
high values for the user/system CPU time consumed in an interval by a given task.
For example:

	GROUP 
	  Catalog = EXT_GROUP|EXC_DEFAULT|EXD_GROUP_TASK_INTERVAL 
	  ITEM 
	    Catalog = EXT_UINT32|EXC_DEFAULT|EXD_TASK_TASKID 
	    Value = 14 
	  ITEM 
	    Catalog = EXT_UINT32|EXC_DEFAULT|EXD_TASK_PROJID 
	    Value = 1000 
	  ITEM 
	    Catalog = EXT_UINT64|EXC_DEFAULT|EXD_TASK_CPU_USER_SEC 
	    Value = 18446744073709551609 

These unrealistic value correspond to the migration of a process from one
task to another, e.g. by newtask(1). This bug occurs when the departing process
has a greater cumulative user/system time then the CPU time consumed by
the remaining processes in the given interval.

[Previous description moved to an attachment]
[ Moved to an attachment ]
Work Around 
N/A
Comments 
N/A