962238 - Display memory usage in Datazilla

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::PerformanceTest, defect)

Description

[Hubert Figuiere [:hub]]

Make sure we do display the memory usage in datazilla.

Comment 1

[Hubert Figuiere [:hub]]

Maybe there is nothing to do but we don't have the jobs running, see blockers.

I just want to make sure we display the info.

Comment 2

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Looking at the memory data and the mozperf.py script, it does not appear that any of this will be submitted without some changes to the latter. At the moment only mozPerfDurations are submitted for each stats/passes entry. We would need to enhance this to push the memory metrics to datazilla.

See https://github.com/mozilla/b2gperf/blob/0.18/b2gperf/mozperf.py#L25

Comment 3

[Jonathan Griffin (:jgriffin)]

The data structure of the JSON looks like:  

15:35:25 {
15:35:25   "stats": {
15:35:25     "suites": 1,
15:35:25     "tests": 1,
15:35:25     "passes": 1,
15:35:25     "pending": 0,
15:35:25     "failures": 0,
15:35:25     "start": "2014-02-28T23:06:48.331Z",
15:35:25     "end": "2014-02-28T23:08:31.946Z",
15:35:25     "duration": 103615,
15:35:25     "application": "email"
15:35:25   },
15:35:25   "failures": [],
15:35:25   "passes": [
15:35:25     {
15:35:25       "title": "startup time ",
15:35:25       "fullTitle": "startup test > email > startup time ",
15:35:25       "duration": 103123,
15:35:25       "mozPerfDurations": [
15:35:25         4107,
15:35:25         4478,
15:35:25         4455,
15:35:25         4979,
15:35:25         5099
15:35:25       ],
15:35:25       "mozPerfDurationsAverage": 4623.6,
15:35:25       "mozPerfMemory": [
15:35:25         {
15:35:25           "app": {
15:35:25             "name": "E-Mail",
15:35:25             "uss": 12.8,
15:35:25             "pss": 17.3,
15:35:25             "rss": 29.9,
15:35:25             "vsize": 68.5
15:35:25           },
15:35:25           "system": {
15:35:25             "name": "b2g",
15:35:25             "uss": 62.5,
15:35:25             "pss": 68,
15:35:25             "rss": 79.5,
15:35:25             "vsize": 190.5
15:35:25           }
15:35:25         },
15:35:25         {
15:35:25           "app": {
15:35:25             "name": "E-Mail",
15:35:25             "uss": 12.7,
15:35:25             "pss": 17.4,
15:35:25             "rss": 30.4,
15:35:25             "vsize": 74.9
15:35:25           },
15:35:25           "system": {
15:35:25             "name": "b2g",
15:35:25             "uss": 64.8,
15:35:25             "pss": 70.2,
15:35:25             "rss": 81.9,
15:35:25             "vsize": 191.5
15:35:25           }
15:35:25         },
15:35:25         {
15:35:25           "app": {
15:35:25             "name": "E-Mail",
15:35:25             "uss": 12.5,
15:35:25             "pss": 17.2,
15:35:25             "rss": 30.2,
15:35:25             "vsize": 74.8
15:35:25           },
15:35:25           "system": {
15:35:25             "name": "b2g",
15:35:25             "uss": 65.5,
15:35:25             "pss": 71,
15:35:25             "rss": 82.6,
15:35:25             "vsize": 188.5
15:35:25           }
15:35:25         },
15:35:25         {
15:35:25           "app": {
15:35:25             "name": "E-Mail",
15:35:25             "uss": 12.7,
15:35:25             "pss": 17.4,
15:35:25             "rss": 30.3,
15:35:25             "vsize": 74.9
15:35:25           },
15:35:25           "system": {
15:35:25             "name": "b2g",
15:35:25             "uss": 63.5,
15:35:25             "pss": 68.9,
15:35:25             "rss": 80.5,
15:35:25             "vsize": 187.5
15:35:25           }
15:35:25         },
15:35:25         {
15:35:25           "app": {
15:35:25             "name": "E-Mail",
15:35:25             "uss": 13.2,
15:35:25             "pss": 17.9,
15:35:25             "rss": 30.8,
15:35:25             "vsize": 74.7
15:35:25           },
15:35:25           "system": {
15:35:25             "name": "b2g",
15:35:25             "uss": 65.3,
15:35:25             "pss": 70.7,
15:35:25             "rss": 82.3,
15:35:25             "vsize": 188.5
15:35:25           }
15:35:25         }
15:35:25       ],
15:35:25       "mozPerfMemoryAverage": {
15:35:25         "app": {
15:35:25           "uss": "12.780",
15:35:25           "pss": "17.440",
15:35:25           "rss": "30.320",
15:35:25           "vsize": "73.560"
15:35:25         },
15:35:25         "system": {
15:35:25           "uss": "64.320",
15:35:25           "pss": "69.760",
15:35:25           "rss": "81.360",
15:35:25           "vsize": "189.300"
15:35:25         }
15:35:25       }
15:35:25     }
15:35:25   ]
15:35:25 },

Comment 4

[Jonathan Griffin (:jgriffin)]

Attachment: Link to Github pull-request: https://github.com/mozilla/b2gperf/pull/42

Comment 5

[Jonathan Griffin (:jgriffin)]

The way we're currently using the datazilla UI needs to be stretched a bit in order to accommodate this memory data, but here's a way we can do it without changes to datazilla.

For each app, a new metric will be created, e.g., "browser_memory".  For each test, we submit data for several different fields per metric:  uss, pss, rss, and vsize (for the app process) and system_uss, system_pss, system_rss, and system_vsize for the b2g process.  Thus, if you click on the "browser_memory" metric in datazilla's lower pane, you should see a graph of all 8 of these metrics for the browser app in the display.  Each app will have its own metric, so we'll have "browser_memory", "calendar_memory", etc.

Comment 6

[Jonathan Griffin (:jgriffin)]

Comment on attachment 8384996 [details] [review]
Link to Github pull-request: https://github.com/mozilla/b2gperf/pull/42

Hub, can you give feedback on the proposed display mechanism in comment #5?

Comment 7

[Hubert Figuiere [:hub]]

Jonathan, I'm not sure I visualize what you are explaining :-(

Comment 8

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Comment on attachment 8384996 [details] [review]
Link to Github pull-request: https://github.com/mozilla/b2gperf/pull/42

r+ with the comment in the pull request addressed.

In order to demonstrate, I pushed to datazilla.allizom.org expecting this to be excluded from the production site. My memory failed me though, as I think this is just staging for the UI...

I'll attach a screenshot of the results and file a bug to remove the noise from the dashboard once this bug is resolved.

Comment 9

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Attachment: Screenshot 2014-03-06 11.13.13.png

Comment 10

[Hubert Figuiere [:hub]]

Comment on attachment 8384996 [details] [review]
Link to Github pull-request: https://github.com/mozilla/b2gperf/pull/42

Output looks good to me, minus the vertical axis that is mislabeled. But that's a detail.

Comment 11

[Jonathan Griffin (:jgriffin)]

Review comment addressed:  https://github.com/mozilla/b2gperf/commit/5b91b333f00139a8ca96981c67c17a4bc59c505e

I'll do a versionbump of b2gperf in a separate bug.

Comment 12

[Dave Hunt [:davehunt] [he/him] ⌚BST]

I've updated Jenkins to use b2gperf 0.20 so we should see some results appearing in datazilla soon.

Comment 13

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Marking as fixed, we now have results in datazilla.