Opened 13 years ago
Last modified 10 years ago
#1097 new task
IZ156: create testsuite test for intermediate accounting records
Reported by: | joga | Owned by: | |
---|---|---|---|
Priority: | high | Milestone: | |
Component: | testsuite | Version: | current |
Severity: | Keywords: | tests | |
Cc: |
Description
[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=156]
Issue #: 156 Platform: All Reporter: joga (joga) Component: testsuite OS: All Subcomponent: tests Version: current CC: None defined Status: NEW Priority: P2 Resolution: Issue type: TASK Target milestone: milestone 1 Assigned to: joga (joga) QA Contact: joga URL: * Summary: create testsuite test for intermediate accounting records Status whiteboard: Attachments: Issue 156 blocks: Votes for issue 156: Vote for this issue Opened: Wed Oct 31 04:39:00 -0700 2007 ------------------------ For long running jobs (over midnight), qmaster is writing intermediate accounting records containing the usage consumed during the last day. The sum of all accounting records is the full job accounting. We need a testsuite test for this feature. It should be implemented in the arco checktree. As waiting for intermediate records being written only at midnight would make the test too long running, we have to instrument sge_qmaster. Currently there is code (a #if 1 and a #if 0) in daemons/qmaster/sge_reporting_qmaster.c which, if changed, makes qmaster write an accounting record every hour. This code should be changed to - write the accounting record even more often, e.g. once a minute - this behavior should be switched on by an environment variable The test has to check: - start_time of intermediate records must be either the job start_time (for the first record), or the endtime of a previous intermediate record. - correctness the accounting information for cpu, mem, io, iow: The sum over all records must match the accounting provided by qacct for the job. - exit_status for intermediate records must be -1 - the query for jobs completed must report the correct number of jobs, not count jobs multiple times, and not count still running jobs - ... When switching on global config, execd_params, SHARETREE_RESERVED_USAGE and ACCT_RESERVED_USAGE, we get some usefull values for cpu, when in addition at job submission we specify -l h_vmem=100M, we also get values for mem, even when using sleeper jobs. ------- Additional comments from joga Wed Nov 21 03:42:46 -0700 2007 ------- I take this IZ.
Note: See
TracTickets for help on using
tickets.