Re: [codesprints] 7.39.1.dev4

Fact-checking myself:

Native : 926.222s

Docker: 1196.506s

(Native without selenium: 607.834s)

So, really, the slowdown is only 30%

RjS

On 11/17/21 8:10 AM, Robert Sparks wrote:
>
> On 11/16/21 9:56 PM, Nick wrote:
>> The Markdown package has been updated today to 3.3.5 and it prevents 
>> the server from starting.
>> Pinning the package version to 3.3.4 in requirements.txt resolves the 
>> issue.
>>
>> @Robert Sparks:
>> By default, Docker for Desktop (mac and Windows) sets a limit on the 
>> number of CPU cores and RAM available to containers.
>> This is likely why you see a performance impact when running the 
>> tests inside the container vs on the host.
>>
>> On Mac, you can increase these settings under Preferences > Advanced, 
>> in Docker for Desktop.
>
> Already done, long ago. The slowdown is, as far as I've been able to 
> discern, in the filesystem sync done into the containers through the 
> VM docker necessarily runs on OSX. (Windows requires one too, of course).
>
> I'm curious what the total test time is on your windows host?
>
> Fwiw, on my 2.4G 8core i9, 64Gb macbook pro, running natively, the 
> test suite takes just under 10 minutes.
>
> I'm currently giving docker 8 cpus, 18Gb ram, and 2Gb swap. There the 
> test suite takes just over 20 minutes.
>
> Incrementing the resources above has not had any appreciable effect.
>
> RjS
>
>> Nick
>>
>> -----Original Message-----
>> From: Lars Eggert <lars@eggert.org>
>> Sent: November 12, 2021 1:37 AM
>> To: Robert Sparks <rjsparks@nostrum.com>
>> Cc: codesprints@ietf.org
>> Subject: Re: [codesprints] 7.39.1.dev4
>>
>> It will be faster - the download time seems to be bound by the 
>> IETF.org pipe currently, whereas I can pull from ghcr.io at tens of 
>> MB/s.
>>
>> -- 
>> Sent from a mobile device; please excuse typos.
>>
>>> On Nov 11, 2021, at 21:30, Robert Sparks <rjsparks@nostrum.com> wrote:
>>>
>>> Btw - we will, when we can, be building these images automatically 
>>> and continuously. The pain of this first 20m build will go away (but 
>>> will be replaced with some initial download time that's on the order 
>>> of the time taken to download the database snapshots).
>>>
>>> RjS
>>>
>>>
>>>> On 11/11/21 1:27 PM, Robert Sparks wrote:
>>>> All -
>>>>
>>>> This update unifies Nick's and Lars' work on the development 
>>>> containerization and shifts us to docker-compose so that we can 
>>>> manage the development database separately from the code.
>>>>
>>>> It also brings in a few changes to the actual code.
>>>>
>>>> Russ - I've tested that this tag's container actually works on os/x 
>>>> up through running the tests.
>>>>
>>>> Note that the instructions for using the container have changed 
>>>> _completely_. Please read through 
>>>> https://trac.ietf.org/trac/ietfdb/browser/tags/dev/7.39.1.dev4/docker/README.md 
>>>> before proceeding.
>>>>
>>>> Consider cleaning out old datatracker docker images and containers.
>>>>
>>>> The first build takes about 20 minutes. After that, getting in and 
>>>> out of the container is very quick.
>>>>
>>>> Tests in the container take 20 minutes to complete on my machine, 
>>>> which is about twice as long as they take natively.
>>>>
>>>> RjS
>>>>
>>>> Since .dev3:
>>>>
>>>>   * Merged in ^/personal/nick/7.39.1.dev3@19629 from 
>>>> nick@staff.ietf.org:
>>>>      feat: Single docker configuration for VS Code and generic dev 
>>>> [19604]
>>>>      fix: remove deprecated files from single docker setup changes 
>>>> [19605]
>>>>      fix: finalize docker setup consolidation [19622]
>>>>      misc: add cleandb script for docker + fix utf8 db charset params
>>>> [19628]
>>>>
>>>>    * Docker related changes: Mark cleandb executable. Clean 
>>>> settings_local
>>>>      and settings_local_sqlitetest of flakes and types issues.
>>>>
>>>>    * Adjust docker volume mounting strategy.
>>>>
>>>>    * Merged in [19621] from rjsparks@nostrum.com:
>>>>      Use rfc8989 for 2022 nomcom eligibility. Fixes #3442.
>>>>
>>>>    * Merged in [19636] from jennifer@painless-security.com:
>>>>      Update groupfeatures_session_purposes field migration to match 
>>>> change
>>>>      to model.
>>>>
>>>>    * Add a command to clean all datatracker related docker 
>>>> resources (from
>>>>      nick@staff.ietf.org)
>>>>
>>>>
>>>>> On 11/9/21 4:22 PM, Robert Sparks wrote:
>>>>> Yes, I know these are coming close together, but a lot is coming 
>>>>> together closely.
>>>>>
>>>>> 7.39.1.dev3 is now available.
>>>>>
>>>>> This brings in a major refactor of the Meeting app to give Purpose 
>>>>> to Sessions, and to remove the old schedule editor.
>>>>>
>>>>> Instructions to get a new working copy are again similar to what's 
>>>>> below.
>>>>>
>>>>> A dump of what the changelog will look like (so far) of the next 
>>>>> release is below. It's going to be a big release.
>>>>>
>>>>> On 11/8/21 2:59 PM, Robert Sparks wrote:
>>>>>> 7.39.1.dev2 is now available. Modify the below instructions 
>>>>>> accordingly.
>>>>>>
>>>>>> RjS
>>>>>>
>>>>>> On 11/2/21 3:46 PM, Robert Sparks wrote:
>>>>>>> I've merged what we've all committed so far, and made a new 
>>>>>>> development tag at 7.39.1.dev1.
>>>>>>>
>>>>>>> You can get a copy of it as follows: In your current checkout 
>>>>>>> top level, run:
>>>>>>>
>>>>>>> $ ./bin/mkdevbranch (your branch root) $  svn co ^/personal/(your
>>>>>>> branch root)/7.39.1.dev1 ../7.39.1.dev1
>>>>>>>
>>>>>>> RjS
>>>>>>>
>>>> _______________________________________________
>>>> codesprints mailing list
>>>> codesprints@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/codesprints
>>> _______________________________________________
>>> codesprints mailing list
>>> codesprints@ietf.org
>>> https://www.ietf.org/mailman/listinfo/codesprints
>>