[OOTB-hive] Docker(/Kubernetes) - My ideas for potential work (re: Daren's mail)
Ole Hejlskov
ole at phpfreak.dk
Fri Dec 7 13:27:13 GMT 2018
Hi all
Forgive me, this will be a wall-of-text long email. TL;DR at the bottom for the lazy.
First of all, I just wanted to recognise the impact the lack of a compatible SDK release has for all of you.
Initially when we released SDK 3.0, it came with a bunch of promises of compatibility. However, this was all before we went into the world of microservices and containers. ACS6 has brought a lot of (good) changes, but unfortunately it has resulted in unfortunate side effects for the SDK.
I am happy to say that we (Alfresco) *are* committed to bring a compatible SDK. We are working out the details on exactly when this might happen, and we hope to share more details soon. There.. I said it. "Soon". I flat out know how this will rub people the wrong way, but that is honestly the best I can do.
OK with all of that out of the way, let's talk details.
We made good progress during the hackathon, and lots of discussions later the approach we took was to spin up the related services (database and search services) as Docker containers.
Now, why did we do this?
Several reasons:
1) Solr isn't a WAR file anymore, it's an executable. We *can* have Maven spin up an external executable (and stop it again) as part of the build lifecycle, but getting this to work across multiple platforms will be a nightmare, if not straight out impossible, unless we leverage profiles (and thereby bring back the dreaded parent pom), which will be hard to maintain. In SDK 3.0 we wrapped all the craziness of dealing with Solr packaging into the Alfresco Maven Plugin. By externalising this into a docker container we removed A LOT of complexity in the plugin.
Solution? Spin it up externally manually, or leverage Docker to automate it.
2) No more H2.
Since ACS no longer leverages Hibernate, we can't control the dialect which enabled us to use H2. There is no way to cheat or work around it, can't be done as is today. ACS will require a database to be running.
Solution? Spin it up externally manually, or leverage Docker to automate it.
OK, so during the hackathon we relied on using the Tomcat Maven Plugin to start the repository itself. This worked at the time, but recent tests have discovered some very, very annoying issues:
a) The Tomcat plugin has not been updated in years. I believe around 5 years.
b) The Tomcat plugin is stuck on Tomcat7
c) The default version shipped with the Tomcat plugin will not load JARs compiled with Java9 (throwing exceptions about module-info.class and stuff). Latest versions of ACS Enterprise relies on some artifacts that contains a module-info.class, so exceptions are all over the place. Without having tested it, I believe the latest Community release will have the same issue
d) You can, in theory, force the Tomcat plugin to use a different version, however you will be limited to Tomcat7 since there are incompatibilities. For the last 5 years this issue has been open (https://issues.apache.org/jira/browse/MTOMCAT-234) which would bring support for Tomcat8, but no resolution or progress made in years
If anyone questions this, I'd be happy to dig up my notes and search histories, there are A LOT of issues with the plugin due it's been neglected for 5 years.
OK, so a lot of issues with the Tomcat plugin. If we bring out the big hammer and force the Tomcat plugin to the latest Tomcat7 release, it will actually start without exceptions... if you are on Java8. Anything higher and it completely breaks apart.. you see where this is going.
This is pretty much where things grinds to a halt.
I no longer believe using an embedded Tomcat like this is viable.
I would be willing to do experiments using the Cargo plugin, but given the direction we will only start to see more smaller services, Transformation service and event gateway just to name a few. It's not impossible to configure all of this without Docker, but it will be very complex and extremely difficult to maintain. (Remember how bad upgrading the between SDK 2.x releases?)
The current design goals is to also leverage Docker to spin up ACS to do integration tests. Ultimately the SDK will consist of an archetype that can compile a JAR and AMP, then leverage Docker to orchestrate the integration tests.
If you do not wish to leverage Docker, there is nothing stopping you from removing the Docker orchestration from the pom in your project and roll a custom setup. This has always been the goal: The archetypes serves as an example that gets you going quickly, but provide enough flexibility to make changes and only use the bits you need for your project.
I assume the next logical question will be how this affects RAD and the overall development workflow. Spinning up ACS as a Docker container sounds like it will be painfully slow, and you have to wait for restarts on every change.
Well, there are different approaches to be taken. We can mount the target directory into the container. We can leverage some stuff from JRebel, or look at how DECVM/Spring Loaded/HotSwap Agent are doing today. I have not personally done experiments in this area yet, so I'd love to hear feedback from people who have experimented with it.
Again, please keep in mind that the archetypes can be tweaked and changed based on your needs. Some people loathe Docker. Some people hate using the embedded Tomcat. The archetypes are meant to give you a good starting point and you can totally tweak it to meet your needs.
As always, we are very open to feedback. I will find time to commit my latest experiments (leveraging Tomcat plugin) to the sdk-4.0 branch ASAP.
TL;DR:
There will be a compatible SDK with ACS 6.0 and 6.1.
The SDK will leverage Docker to spin up services such as Postgres and Search Services, and ultimately the repository itself.
/Ole
> On 7 Dec 2018, at 11.14, Douglas C. R. Paes <douglascrp at gmail.com> wrote:
>
> Hello Jeff.
>
> The idea is not to force the Docker use, but to use it to make things easier. This is why Docker is used, to start PostgreSQL and Search Services only.
> For Alfresco and Share, there is nothing different, as the war files are used as always.
> But those are optional, and one can have then running locally using traditional methods.
>
>
> Douglas C. R. Paes
>
> "Done is better than perfect"
>
>
> Em qui, 6 de dez de 2018 às 23:45, Jeff Potts <jeffpotts01 at gmail.com <mailto:jeffpotts01 at gmail.com>> escreveu:
> Why would I want the SDK to use a Docker image?
>
> My opinion is that it should continue to use the WAR as it has previously. Otherwise, as you point out, the developer has to add technology they may not have a need for.
>
> Jeff
>
> On Thu, Dec 6, 2018 at 11:36 AM Douglas C. R. Paes <douglascrp at gmail.com <mailto:douglascrp at gmail.com>> wrote:
> Hello everyone.
>
> Because of the lack of support from Alfresco in the SDK space, and because I want to start using the Alfresco 6 version, I decided to check on the current status of the SDK project, and I took the work started by Ole on the last Hackathon, and tried to make it work.
>
> The result of my effort so far is available at https://github.com/douglascrp/alfresco-sdk/tree/sdk-4.0 <https://github.com/douglascrp/alfresco-sdk/tree/sdk-4.0>
>
> Important... I have been working only with alfresco-platform-jar-archetype, as I usually don't use the AIO archetype for anything.
>
> When you create the project using the alfresco-platform-jar-archetype, the project will be configured to start both PostgreSQL and Alfresco Search Service as Docker containers, but the project itself will be executed using the tomcat plugin, just like it was done in the SDK version 3.0.1.
>
> As this relies on the official Alfresco Docker image for Alfresco Search Service, anyone running it will have to perform the configuration described by Angel at https://github.com/keensoft/alfresco-docker-template/tree/master/templates/201806-GA#how-to-use-this-composition <https://github.com/keensoft/alfresco-docker-template/tree/master/templates/201806-GA#how-to-use-this-composition>, which is to set the owner of the .data/solr-data.
> On Ubuntu: "sudo chown -R 1000:1000 data/solr-data"
>
> There are still some problems with activiti, and because of that, I had to configure the engline to be disabled by default.
>
> Feel free to test it and tell me what you think.
>
> Douglas C. R. Paes
>
> "Done is better than perfect"
>
>
> Em seg, 26 de nov de 2018 às 10:16, Angel Borroy <angel.borroy at keensoft.es <mailto:angel.borroy at keensoft.es>> escreveu:
> You are right, Axel.
>
> Sorry for trolling your initial (and great) initiative.
>
> Hoping people will fallback to main subject of this thread...
>
> Angel Borroy
> keensoft a UST Global company
>
> CONFIDENCIALIDAD:
>
> La presente comunicación y, en su caso, los ficheros que lleve adjuntos, pertenecen exclusivamente a las personas a las que va dirigido y puede contener información confidencial. Si usted no es el destinatario de este mensaje (o la persona responsable de su entrega), considérese advertido de que lo ha recibido por error y que cualquier uso, difusión, reenvío o copia están prohibidos legalmente. Si ha recibido este mensaje por error, por favor notifíquelo al remitente y proceda a destruirlo inmediatamente.
>
> This message and the attached documents may contain privileged/confidential information and intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient (or responsible for delivery of the message to such a person) be advised that you have received this message in error and that any use, dissemination, forwarding, printing or copying of this e-mail is strictly prohibited. If you have received this message in error please notify it to the sender and destroy it immediately.
>
>
> On 26 November 2018 at 13:14:33, Axel Faust (axel.faust.g at googlemail.com <mailto:axel.faust.g at googlemail.com>) wrote:
>
>> This is only using a test / RC (release candidate) state of 6.1 EE - if you look in artifacts.alfresco.com <http://artifacts.alfresco.com/> (I can via one of my customers) you'll see there is no EE GA yet.
>>
>> Haven't we all learned by now not to trust Alfresco marketing (e.g. what's written on the homepage) when it comes to 100% correctly / transparently representing facts? CE vs EE comparison has been (naturally) biased forever, and since Alfresco was planning a marketing push for November it is also understandable that they might want to highlight the next release, even if it may not yet be fully ready...
>>
>> Of course as always, advance / roadmap communication by Alfresco could be better to avoid any such confusion.
>>
>> On 26/11/2018 12:56, Angel Borroy wrote:
>>> At least, in my case, comes from here:
>>>
>>> <part1.00570E7E.7B019451 at googlemail>
>>>
>>> This looks like a public ACS EE 6.1 release...
>>>
>>> Angel Borroy
>>> keensoft a UST Global company
>>>
>>> email: angel.borroy at keensoft.es <mailto:angel.borroy at keensoft.es>
>>> web: http://www.keensoft.es <http://www.keensoft.es/>
>>> móvil: +34 655 47 47 55 <tel://+34%20655%2047%2047%2055>
>>> CONFIDENCIALIDAD:
>>>
>>> La presente comunicación y, en su caso, los ficheros que lleve adjuntos, pertenecen exclusivamente a las personas a las que va dirigido y puede contener información confidencial. Si usted no es el destinatario de este mensaje (o la persona responsable de su entrega), considérese advertido de que lo ha recibido por error y que cualquier uso, difusión, reenvío o copia están prohibidos legalmente. Si ha recibido este mensaje por error, por favor notifíquelo al remitente y proceda a destruirlo inmediatamente.
>>>
>>> This message and the attached documents may contain privileged/confidential information and intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient (or responsible for delivery of the message to such a person) be advised that you have received this message in error and that any use, dissemination, forwarding, printing or copying of this e-mail is strictly prohibited. If you have received this message in error please notify it to the sender and destroy it immediately.
>>>
>>>
>>> On 26 November 2018 at 12:54:48, Ole Hejlskov (ole at phpfreak.dk <mailto:ole at phpfreak.dk>) wrote:
>>>
>>>> Hi all
>>>>
>>>> I’m not sure where the confusion started, but I’ve just verified internally that ACS 6.1 has entered the regression testing phase. Once this completes there will be a ACS 6.1 GA release for Community, followed by 6.1 EE. The only thing that isn’t clear to me at this stage, is wether 6.1 EE comes before 6.1 CE (GA) or vice-versa.
>>>>
>>>> Hope this clears things up.
>>>>
>>>> /Ole
>>>>
>>>>> On 26 Nov 2018, at 11.33, Francesco Corti <fcorti at gmail.com <mailto:fcorti at gmail.com>> wrote:
>>>>>
>>>>> Good morning guys,
>>>>>
>>>>> I'm afraid there is something to adjust.
>>>>> ACS 6.1 Community Edition won't be a GA release.
>>>>> I think John mentioned the EE release (that will be GA by default).
>>>>> Just a heads up.
>>>>> In every case, the initiative continue to be a good idea.
>>>>>
>>>>> Cheers,
>>>>>
>>>>> -F
>>>>>
>>>>>
>>>>>
>>>>> Il giorno lun 26 nov 2018 alle ore 10:29 Angel Borroy <angel.borroy at keensoft.es <mailto:angel.borroy at keensoft.es>> ha scritto:
>>>>> Hi,
>>>>>
>>>>> Sorry for being so vague. I have no more information than you, as UST Global is not still an Alfresco partner. We are applying from June but it looks like Alfresco has plenty of skilled partners in Spain and there is no room for one more. Two weeks ago I started an Enterprise project in collaboration with a local partner, as Alfresco forbid UST Global to work with Community (despite we are not partners nowadays). I’m starting to be very concerned about Alfresco Sales ecosystem, but probably it’s only my pov.
>>>>>
>>>>> Anyway, you are right, John Knowles said that they will publish an ACS Community release in GA status. But, afaik, this is the first Community release released after the Enterprise one. Error or strategy? Who knows…
>>>>>
>>>>> Let’s focus on this initiative and let’s keep apart all my bad feelings about the life in general.
>>>>>
>>>>> Best,
>>>>>
>>>>> Angel Borroy
>>>>> keensoft a UST Global company
>>>>>
>>>>> email: angel.borroy at keensoft.es <mailto:angel.borroy at keensoft.es>
>>>>> web: http://www.keensoft.es <http://www.keensoft.es/>
>>>>> móvil: +34 655 47 47 55 <>
>>>>> CONFIDENCIALIDAD:
>>>>>
>>>>> La presente comunicación y, en su caso, los ficheros que lleve adjuntos, pertenecen exclusivamente a las personas a las que va dirigido y puede contener información confidencial. Si usted no es el destinatario de este mensaje (o la persona responsable de su entrega), considérese advertido de que lo ha recibido por error y que cualquier uso, difusión, reenvío o copia están prohibidos legalmente. Si ha recibido este mensaje por error, por favor notifíquelo al remitente y proceda a destruirlo inmediatamente.
>>>>>
>>>>> This message and the attached documents may contain privileged/confidential information and intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient (or responsible for delivery of the message to such a person) be advised that you have received this message in error and that any use, dissemination, forwarding, printing or copying of this e-mail is strictly prohibited. If you have received this message in error please notify it to the sender and destroy it immediately.
>>>>>
>>>>>
>>>>> On 26 November 2018 at 09:43:45, Axel Faust (axel.faust.g at googlemail.com <mailto:axel.faust.g at googlemail.com>) wrote:
>>>>>
>>>>>> Hello Angel,
>>>>>>
>>>>>> since you now are a partner, you may have some information I don't get to see - but last time John Knowles was on IRC (~2 weeks ago) he indicated that the next ACS Community release would be a GA, so ACS 6.1 would have a Community equivalent.
>>>>>>
>>>>>> Regards, Axel
>>>>>>
>>>>>> On 26/11/2018 09:36, Angel Borroy wrote:
>>>>>>> Nice idea, Axel.
>>>>>>>
>>>>>>> I’m open to collaborate, despite my basic skills on Docker, k8s and many other relevant software involved in the “new” Alfresco.
>>>>>>>
>>>>>>> Now that Alfresco is ignoring deliberatively the Community (ACS 6.1 will be published without having a Community equivalent), it’s time to act. Probably providing a deployment and developing environment for everyone is the right approach.
>>>>>>>
>>>>>>> I’m not able to organise this initiative, but let me know if I can help in any way.
>>>>>>>
>>>>>>> Best,
>>>>>>>
>>>>>>> Angel Borroy
>>>>>>> keensoft a UST Global company
>>>>>>>
>>>>>>> email: angel.borroy at keensoft.es <mailto:angel.borroy at keensoft.es>
>>>>>>> web: http://www.keensoft.es <http://www.keensoft.es/>
>>>>>>> móvil: +34 655 47 47 55 <tel://+34%20655%2047%2047%2055>
>>>>>>> CONFIDENCIALIDAD:
>>>>>>>
>>>>>>> La presente comunicación y, en su caso, los ficheros que lleve adjuntos, pertenecen exclusivamente a las personas a las que va dirigido y puede contener información confidencial. Si usted no es el destinatario de este mensaje (o la persona responsable de su entrega), considérese advertido de que lo ha recibido por error y que cualquier uso, difusión, reenvío o copia están prohibidos legalmente. Si ha recibido este mensaje por error, por favor notifíquelo al remitente y proceda a destruirlo inmediatamente.
>>>>>>>
>>>>>>> This message and the attached documents may contain privileged/confidential information and intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient (or responsible for delivery of the message to such a person) be advised that you have received this message in error and that any use, dissemination, forwarding, printing or copying of this e-mail is strictly prohibited. If you have received this message in error please notify it to the sender and destroy it immediately.
>>>>>>>
>>>>>>>
>>>>>>> On 25 November 2018 at 22:57:06, Axel Faust (axel.faust.g at googlemail.com <mailto:axel.faust.g at googlemail.com>) wrote:
>>>>>>>
>>>>>>>> Hello guys,
>>>>>>>>
>>>>>>>> TL;DR: Instead of revisiting our previous pre-bundled Community build
>>>>>>>> idea, maybe more short-term / easier to realize value could be gained by
>>>>>>>> collaborating on guides / documentation / examples regarding Docker
>>>>>>>> deployments to help community members make the switch, AND/OR create a
>>>>>>>> standard build setup (in lieu of an updated Alfresco SDK that may never
>>>>>>>> arrive) incorporating the new deployment approach and achieving full
>>>>>>>> compatibility with the most recent Alfresco releases.
>>>>>>>>
>>>>>>>> This mail is a revisit of a mail Daren sent in the middle of the thread
>>>>>>>> for volunteers for the DevCon talk submission review ([1]). I don't know
>>>>>>>> how many actually read the mail considering the misleading subject line.
>>>>>>>> In it he asked if there is interest in revisiting our previous work
>>>>>>>> around a pre-bundled Alfresco Community build (at the time using puppet)
>>>>>>>> and maybe adapted / updating that to be in line with the current Docker
>>>>>>>> / Kubernetes approach.
>>>>>>>>
>>>>>>>> In the last two months I have had the (dis-)pleasure to familiarize
>>>>>>>> myself with the current state of the Alfresco Docker and Helm charts /
>>>>>>>> Kubernetes work in preparation for a training at an Alfresco partner.
>>>>>>>> Anyone following me on Twitter might have seen a rant or two. It
>>>>>>>> certainly is not the most intuitive and not helped by the fact that
>>>>>>>> Alfresco tends to break stuff shortly after providing an initially
>>>>>>>> working state.
>>>>>>>>
>>>>>>>> Given how tricky it can be for community members / customers to switch
>>>>>>>> over to the new approach without any prior experience, I would say that
>>>>>>>> any guides / documentation that we as the Order could collaboratively
>>>>>>>> put together would be much more valuable/needed at this point than
>>>>>>>> simply revisiting the pre-bundled build. As both the Docker Compose
>>>>>>>> scripts and Helm charts from Alfresco should only be considered as a
>>>>>>>> starting point / reference rather than a ready-to-use setup, we might
>>>>>>>> also want to look into providing improved variants for these (Helm may
>>>>>>>> be a lower priority as it/Kubernetes likely is not relevant for most
>>>>>>>> community users). I know that Angel has already done work in the area of
>>>>>>>> an improved compose reference ([2]) - maybe that could be a starting
>>>>>>>> point to flesh out / improve upon...
>>>>>>>>
>>>>>>>> In my training I have also integrated the new Docker images as a runtime
>>>>>>>> environment for ("not-really-rapid") application development in my Maven
>>>>>>>> setup, and also used it for the integration tests therein. In another,
>>>>>>>> personal project I have started to use Docker containers even for my
>>>>>>>> regular unit tests.
>>>>>>>> Alfresco has still not released a SDK natively compatible with Alfresco
>>>>>>>> 6, and given their long-term goal of moving customers/users off of
>>>>>>>> in-process extensions may likely not provide an updated version anytime
>>>>>>>> soon. I know most people in the Order have sufficient skill to adapt the
>>>>>>>> current SDK to work with 6, but that is likely not true for most
>>>>>>>> community users out there. So maybe we as the Order should look into
>>>>>>>> providing a reference build setup of our own, using the new
>>>>>>>> incorporating the new Docker approach as best as possible. And maybe
>>>>>>>> such work will naturally lead us to general improvements / definition of
>>>>>>>> a base image e.g. with regards to image size, code+test round-trip time
>>>>>>>> etc...
>>>>>>>>
>>>>>>>> Regards
>>>>>>>>
>>>>>>>> Axel
>>>>>>>>
>>>>>>>>
>>>>>>>> [1]
>>>>>>>> http://lists.xtreamlab.net/pipermail/ootb-hive/2018-September/001028.html <http://lists.xtreamlab.net/pipermail/ootb-hive/2018-September/001028.html>
>>>>>>>>
>>>>>>>> [2] https://github.com/keensoft/docker-alfresco <https://github.com/keensoft/docker-alfresco>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> OOTB-hive mailing list
>>>>>>>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>>>>>>>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
>>>>> _______________________________________________
>>>>> OOTB-hive mailing list
>>>>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>>>>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
>>>>> _______________________________________________
>>>>> OOTB-hive mailing list
>>>>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>>>>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
>>>>
>>>> _______________________________________________
>>>> OOTB-hive mailing list
>>>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>>>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
>>>
>>>
>>> _______________________________________________
>>> OOTB-hive mailing list
>>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
>> _______________________________________________
>> OOTB-hive mailing list
>> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
>> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
> _______________________________________________
> OOTB-hive mailing list
> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
> _______________________________________________
> OOTB-hive mailing list
> OOTB-hive at lists.xtreamlab.net <mailto:OOTB-hive at lists.xtreamlab.net>
> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive <https://lists.xtreamlab.net/mailman/listinfo/ootb-hive>
> _______________________________________________
> OOTB-hive mailing list
> OOTB-hive at lists.xtreamlab.net
> https://lists.xtreamlab.net/mailman/listinfo/ootb-hive
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xtreamlab.net/pipermail/ootb-hive/attachments/20181207/c3511871/attachment-0001.html>
More information about the OOTB-hive
mailing list