Age | Commit message (Collapse) | Author | Files | Lines |
|
Changes the overall value of all metrics to be based on the
calculated steady state values instead of the average for the
entire run.
Change-Id: I121929d5fe2dd43df7f289b82e9f5291c9ea9aab
JIRA: STORPERF-107
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Changes the event notification logic inside the FIO invoker
so that it no longer publishes events after termination.
Prevents false reports after steady state has been detected.
Change-Id: I694f77b6493b88820fe4f4cc7f634e3e62c45a9a
JIRA: STORPERF-105
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Changes the data series from using seconds to samples
so that slope is not artificially flattened.
Change-Id: Idf87926a47c2ba67e66e2254d3572adad7a81b44
JIRA: STORPERF-106
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Records the URL returned from testresults db.
Changes the URL ref from localhost to external and reports in
daily job.
Change-Id: I0068ea963671fb183779ac20592ba6933647eea0
JIRA: STORPERF-104
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Changes the key names sent to the testresults db to align with other
projects.
Defines a clear PASS/FAIL based on the ability to find steady state for
all runs. If one run failed to achieve steady state, run gets labelled
as FAIL.
Change-Id: I73143779453689eea93829079388ed8035134aa3
JIRA: STORPERF-100
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Fixes a problem where FIO does not terminate by scheduling a
second killall if we get a specific message back from FIO
stderr.
Introduces a new flavor for StorPerf that has a little more
memory as larger memory maps for duplicate blocks sometimes
caused out of memory killer to be invoked.
Change-Id: I06856561ad73fef582a81d4136a36a1bea47654a
JIRA: STORPERF-99
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Change reporting so that multiple workloads in one job execution can
be reported instead of overwriting the previous value.
Change the daily job to use a single, multiple workload run.
Change-Id: I8e350350ae13d2272b584af7a60ad269de160587
JIRA: STORPERF-98
Signed-off-by: mbeierl <mark.beierl@dell.com>
|
|
Detection of steady state after 10+ samples of data
Change-Id: I29368b06188c6370d17b3d567fece6486d171235
JIRA: STORPERF-72 STORPERF-73
Signed-off-by: Mark Beierl <mark.beierl@dell.com>
|
|
Break out test db interaction into new module and
make the push event driven instead of the sleep
that was there before.
Change-Id: I9485aba1405f6c3b4ee5000168fbc037efa87c81
JIRA: STORPERF-90
Signed-off-by: Mark Beierl <mark.beierl@dell.com>
|
|
Module that allows passing of a gate only once all other
peers have also reported in within a specified time period
Change-Id: If4baf1d4377026c7833a6f30bfc2e36698f675e8
JIRA: STORPERF-71
Signed-off-by: Mark Beierl <mark.beierl@dell.com>
|
|
Test and fix for error in slope
STORPERF-74
Change-Id: I17a8e0e1e588deb875ecace85290f92167df1d5e
Signed-off-by: Mark Beierl <mark.beierl@emc.com>
|
|
Added a Steady State Detection module containing a steady_state(data_series)
function that is able to return a boolean indicating wether or not steady
state is reached with the data_series being passed.
This module requires a data_treatment(data_series) and an average(data_series)
modules that have been added in this commit as well. The data treatment function
aims at formatting the data series that is passed to the high level steady_state
function to reach the requirement of each sub-module (slope, average and range).
Modified the Slope and Range functions so they return None when passed an empty
data series instead of 0 which was wrong. Modified the corresponding test cases.
Modified the math_range_test.py file to fix a bug in the 2 last tests.
Change-Id: I9c3854cb0a21cc37b0787b8afca0821eefaa203d
JIRA: STORPERF-60
JIRA: STORPERF-59
JIRA: STORPERF-61
JIRA: STORPERF-62
Signed-off-by: Tim Rault <tim.rault@cengn.ca>
|
|
Moving the test files into their own top-level
directory to keep things clean
Change-Id: Ic50b881045bc59b003807923424345b335dd5c95
Signed-off-by: Mark Beierl <mark.beierl@emc.com>
|