-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/44229/#review123800
-----------------------------------------------------------




support/test-upgrade.py (line 24)
<https://reviews.apache.org/r/44229/#comment186069>

    Can you document what this function does and returns?



support/test-upgrade.py (line 42)
<https://reviews.apache.org/r/44229/#comment186076>

    credfile is unused?



support/test-upgrade.py (line 47)
<https://reviews.apache.org/r/44229/#comment186064>

    2 blank lines.



support/test-upgrade.py (line 51)
<https://reviews.apache.org/r/44229/#comment186077>

    credfile is unused?



support/test-upgrade.py (line 59)
<https://reviews.apache.org/r/44229/#comment186070>

    s/insteance/instance/



support/test-upgrade.py (line 62)
<https://reviews.apache.org/r/44229/#comment186078>

    credfile is unused?



support/test-upgrade.py (line 63)
<https://reviews.apache.org/r/44229/#comment186072>

    s/persistent-volume-framework/test-framework/ ?
    
    actuallly we should be testing the long-lived-framework instead of 
test-framework. but that needs updating the long lived framework to abort on 
un-expected state changes (e.g., TASK_FAILED received instead of TASK_FINISHED).



support/test-upgrade.py (line 66)
<https://reviews.apache.org/r/44229/#comment186063>

    2 blank lines.



support/test-upgrade.py (line 67)
<https://reviews.apache.org/r/44229/#comment186066>

    s/numbers//
    
    s/tested/built/ ?



support/test-upgrade.py (line 80)
<https://reviews.apache.org/r/44229/#comment186062>

    2 blank lines.



support/test-upgrade.py (line 102)
<https://reviews.apache.org/r/44229/#comment186065>

    s/tested/built/ ?



support/test-upgrade.py (lines 110 - 113)
<https://reviews.apache.org/r/44229/#comment186079>

    seems unused. just kill this for now.



support/test-upgrade.py (lines 118 - 125)
<https://reviews.apache.org/r/44229/#comment186084>

    The upgrade flow we want to test is
    
    scheduler (old) <-> master (old) <-> slave (old)
    scheduler (old) <-> master (new) <-> slave (old)
    scheduler (old) <-> master (new) <-> slave (new)
    scheduler (new) <-> master (new) <-> slave (new)



support/test-upgrade.py (lines 153 - 154)
<https://reviews.apache.org/r/44229/#comment186075>

    if prev_framework.sleep(10) != 0:
      ...



support/test-upgrade.py (line 158)
<https://reviews.apache.org/r/44229/#comment186082>

    Lets add a test for upgrading master first. This is what typically 
operators do.



support/test-upgrade.py (lines 174 - 175)
<https://reviews.apache.org/r/44229/#comment186080>

    no need for prev_framework_result temp variable?



support/test-upgrade.py (lines 188 - 189)
<https://reviews.apache.org/r/44229/#comment186081>

    no need for next_famework_result.



support/test-upgrade.py (lines 197 - 228)
<https://reviews.apache.org/r/44229/#comment186083>

    not sure, why we do it this way. just kill this.


- Vinod Kone


On March 16, 2016, 12:18 a.m., Greg Mann wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/44229/
> -----------------------------------------------------------
> 
> (Updated March 16, 2016, 12:18 a.m.)
> 
> 
> Review request for mesos, Niklas Nielsen and Vinod Kone.
> 
> 
> Bugs: MESOS-2372
>     https://issues.apache.org/jira/browse/MESOS-2372
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Added upgrade testing script.
> 
> 
> Diffs
> -----
> 
>   support/test-upgrade.py PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/44229/diff/
> 
> 
> Testing
> -------
> 
> This script is copied from @nnielsen, original patch can be found here: 
> https://reviews.apache.org/r/31645/
> 
> To test, build two different versions of Mesos in two different directories. 
> Then do:
> `support/test-upgrade.py --prev=/path/to/earlier/version 
> --next=/path/to/later/version`
> 
> 
> Thanks,
> 
> Greg Mann
> 
>

Reply via email to