Project

General

Profile

Actions

Bug #584

open

Stats server doesn't show failed launches correctly

Added by Evgeny Novikov almost 14 years ago. Updated almost 14 years ago.

Status:
Open
Priority:
Normal
Category:
Statistics server
Start date:
11/25/2010
Due date:
% Done:

0%

Estimated time:
Detected in build:
48fd669
Platform:
Published in build:

Description

For example, when build command extractor fails many times on the same kernel, stats server joins all such failed launches together since they all have the same 'statistics key', like ('linux-2.6.31.6', NULL, NULL, NULL) for 'Unknown' page:
http://10.10.2.59:8999/stats/index/name/ldv_qa_night_res/user/qarobot/host/10.10.2.59/password/12345/page/Unknown/value/62/Task+id/31/Task+description/2010-11-25-00-00-02/Environment+version/linux-2.6.31.6/Rule+name/__NULL

This leads that many different problems are joined together and their descriptions do the same (but mysql chooses just one of appropriate descriptions), so we can't analyze them.
The problem is rather artificial, since one task can't contain more then one build fail per a kernel. But we can join a lot of tasks together and then one task can contain a lot of build fails per a kernel. So if we see on unjoined tasks we can analyze all unknowns:
http://10.10.2.59:8999/stats/index/name/ldv_qa_night/host/10.10.2.59/user/qarobot/password/12345

I think that 2 activities should be done:
  • First of all checking that stats server tries to group unjoinable values together due to ununique statistics key (like descriptions above). I've thought about this even on the early development stage and add special flag that says what field requires unique key (it isn't used now). Also I think that it isn't trivial operation and may consume a lot of time (or may be there is appropriate SQL decision exists...) so such check must be optional.
  • Extend statistics key (e.g. add driver name for Unknown page) to pass the check described above and show all correctly.

Related issues 1 (0 open1 closed)

Related to Linux Driver Verification - Bug #587: Strange changes in general regression test setClosedAlexandr Strakh11/26/2010

Actions
Actions #1

Updated by Evgeny Novikov almost 14 years ago

  • Assignee set to Evgeny Novikov
  • Priority changed from Normal to High

At least the second should be done to analyze changes in regression test set that has high priority.

Actions #2

Updated by Evgeny Novikov almost 14 years ago

  • Status changed from New to Open

Rather important for coming huge verification and comparison. May be should be done in January.

Actions #3

Updated by Evgeny Novikov almost 14 years ago

I mark corresponding task through QA night tests results.

Actions #4

Updated by Pavel Shved almost 14 years ago

  • Priority changed from High to Normal

We decided to postpone this.

Actions

Also available in: Atom PDF