5:10 a.m.
was extolling the virtues of podman to a colleague, who seemed
interested but asked what the podman-based alternative was for "docker
compose". rather than mislead her, i thought i would ask the experts.
thoughts?
rday
--
========================================================================
Robert P. J. Day Ottawa, Ontario, CANADA
http://crashcourse.ca
Twitter: http://twitter.com/rpjday
LinkedIn: http://ca.linkedin.com/in/rpjday
========================================================================
6:09 a.m.
Hi Robert,
I'm not an expert at all. But I had this question a few days ago.
The most common tool seems to be
https://github.com/containers/podman-compose.
Unfortunately, in my case it was not able to replace docker's compose
completely¹.
But podman-compose has a option to trace what it is trying to do. This
helped me, to write a simple script, which is doing the same job as the
initial composer file.
Uwe
¹) Sorry I haven't documented the problem well.
As far as I remember, the problem was that composer creates/runs
standalone containers with different IPs and an implicit DNS service
(container name -> ip). podman-compose creates a pod, which means all
containers in this pod share the IP 'localhost'. Im my case this was a
problem, because the application on one container was looking for
resources on the host 'mysql'.
6:41 a.m.
Depending on the complexity of the task, podman generate kube and
podman play kube may also be an answer.
https://developers.redhat.com/blog/2019/01/29/podman-kubernetes-yaml/
While most of the article talks about replaying on a kube-basaed
runtime, one of the last paragraphs talks about replaying locally.
On Fri, 2019-08-30 at 13:09 +0200, Uwe Reh wrote:
Hi Robert,
I'm not an expert at all. But I had this question a few days ago.
The most common tool seems to be
https://github.com/containers/podman-compose.
Unfortunately, in my case it was not able to replace docker's compose
completely¹.
But podman-compose has a option to trace what it is trying to do.
This
helped me, to write a simple script, which is doing the same job as
the
initial composer file.
Uwe
¹) Sorry I haven't documented the problem well.
As far as I remember, the problem was that composer creates/runs
standalone containers with different IPs and an implicit DNS service
(container name -> ip). podman-compose creates a pod, which means all
containers in this pod share the IP 'localhost'. Im my case this was
a
problem, because the application on one container was looking for
resources on the host 'mysql'.
_______________________________________________
Podman mailing list -- podman(a)lists.podman.io
To unsubscribe send an email to podman-leave(a)lists.podman.io
9:19 a.m.
Brent is also working on implementing a strategy so that containers can
communicate with each other via Names. We are looking at enhancing CNI
to support running a DNSMasq/per network device and registering the
containers hostnames with it.
On 8/30/19 7:41 AM, Brent Baude wrote:
Depending on the complexity of the task, podman generate kube and
podman play kube may also be an answer.
https://developers.redhat.com/blog/2019/01/29/podman-kubernetes-yaml/
While most of the article talks about replaying on a kube-basaed
runtime, one of the last paragraphs talks about replaying locally.
On Fri, 2019-08-30 at 13:09 +0200, Uwe Reh wrote:
> Hi Robert,
>
> I'm not an expert at all. But I had this question a few days ago.
>
> The most common tool seems to be
> https://github.com/containers/podman-compose.
>
> Unfortunately, in my case it was not able to replace docker's compose
> completely¹.
> But podman-compose has a option to trace what it is trying to do.
> This
> helped me, to write a simple script, which is doing the same job as
> the
> initial composer file.
>
> Uwe
>
>
> ¹) Sorry I haven't documented the problem well.
> As far as I remember, the problem was that composer creates/runs
> standalone containers with different IPs and an implicit DNS service
> (container name -> ip). podman-compose creates a pod, which means all
> containers in this pod share the IP 'localhost'. Im my case this was
> a
> problem, because the application on one container was looking for
> resources on the host 'mysql'.
>
>
>
> _______________________________________________
> Podman mailing list -- podman(a)lists.podman.io
> To unsubscribe send an email to podman-leave(a)lists.podman.io
_______________________________________________
Podman mailing list -- podman(a)lists.podman.io
To unsubscribe send an email to podman-leave(a)lists.podman.io
3:09 a.m.
Hi Brent,
do you really think, that your article is an appropriate answer to the
initial question? I suppose not.
Surely you've mad a good job for people which are familiar with
Kubernets. For beginners like Robert and me, it's some kind of chinese.
Just try to imaging:
- you are relatively new in the area of containers.
- you have heard just enough about Kubernets, to realize that it's out
of scope for your needs.
- you have read about Kubernet's 'kompose' and have seen your article.
A beginner like me, may think, "Ok, sounds reasonable".
But this leads into an dead end. 'kompose' creates a lot of nice
detailed configuration files, but none of them seems to be suitable for
podman...
Without any experience or deeper knowledge about Kubernet, I resigned.
Uwe
PS.
Maybe, you could write a detailed howto for the new podman pages.
Am 30.08.19 um 13:41 schrieb Brent Baude:
> Depending on the complexity of the task, podman generate kube and
> podman play kube may also be an answer.
>
> https://developers.redhat.com/blog/2019/01/29/podman-kubernetes-yaml/
>
> While most of the article talks about replaying on a kube-basaed
> runtime, one of the last paragraphs talks about replaying locally.
>
> On Fri, 2019-08-30 at 13:09 +0200, Uwe Reh wrote:
>> Hi Robert,
>>
>> I'm not an expert at all. But I had this question a few days ago.
>>
>> The most common tool seems to be
>> https://github.com/containers/podman-compose.
>>
>> Unfortunately, in my case it was not able to replace docker's compose
>> completely¹.
>> But podman-compose has a option to trace what it is trying to do.
>> This
>> helped me, to write a simple script, which is doing the same job as
>> the
>> initial composer file.
>>
>> Uwe
>>
>>
>> ¹) Sorry I haven't documented the problem well.
>> As far as I remember, the problem was that composer creates/runs
>> standalone containers with different IPs and an implicit DNS service
>> (container name -> ip). podman-compose creates a pod, which means all
>> containers in this pod share the IP 'localhost'. Im my case this was
>> a
>> problem, because the application on one container was looking for
>> resources on the host 'mysql'.
>>
>>
>>
>> _______________________________________________
>> Podman mailing list -- podman(a)lists.podman.io
>> To unsubscribe send an email to podman-leave(a)lists.podman.io
> _______________________________________________
> Podman mailing list -- podman(a)lists.podman.io
> To unsubscribe send an email to podman-leave(a)lists.podman.io
>
6:04 a.m.
Uwe,
The opensource community is responding to the need for podman-compose in
github.com/containers/podman-compose.
The issue with docker-compose is it is huge and we don't have the
engineering staff to work and respond to all requests for support on
compose. Our team is goaled at supporting the traditional container
work loads on a single node. And helping people move to the multi-node
containers via Kubernetes and OpenShift.
We work with the community to support other features like compose and
would not be against adding features provided by the community to better
integrate podman into the compose world, but we don't have the time or
staff to do the work ourselves.
As far as documenting how to move a traditional container workload to
Kubernetes, we would be interested in this.
On 9/2/19 4:09 AM, Uwe Reh wrote:
Hi Brent,
do you really think, that your article is an appropriate answer to the
initial question? I suppose not.
Surely you've mad a good job for people which are familiar with
Kubernets. For beginners like Robert and me, it's some kind of chinese.
Just try to imaging:
- you are relatively new in the area of containers.
- you have heard just enough about Kubernets, to realize that it's out
of scope for your needs.
- you have read about Kubernet's 'kompose' and have seen your article.
A beginner like me, may think, "Ok, sounds reasonable".
But this leads into an dead end. 'kompose' creates a lot of nice
detailed configuration files, but none of them seems to be suitable for
podman...
Without any experience or deeper knowledge about Kubernet, I resigned.
Uwe
PS.
Maybe, you could write a detailed howto for the new podman pages.
Am 30.08.19 um 13:41 schrieb Brent Baude:
> Depending on the complexity of the task, podman generate kube and
> podman play kube may also be an answer.
>
> https://developers.redhat.com/blog/2019/01/29/podman-kubernetes-yaml/
>
> While most of the article talks about replaying on a kube-basaed
> runtime, one of the last paragraphs talks about replaying locally.
>
> On Fri, 2019-08-30 at 13:09 +0200, Uwe Reh wrote:
>> Hi Robert,
>>
>> I'm not an expert at all. But I had this question a few days ago.
>>
>> The most common tool seems to be
>> https://github.com/containers/podman-compose.
>>
>> Unfortunately, in my case it was not able to replace docker's compose
>> completely¹.
>> But podman-compose has a option to trace what it is trying to do.
>> This
>> helped me, to write a simple script, which is doing the same job as
>> the
>> initial composer file.
>>
>> Uwe
>>
>>
>> ¹) Sorry I haven't documented the problem well.
>> As far as I remember, the problem was that composer creates/runs
>> standalone containers with different IPs and an implicit DNS service
>> (container name -> ip). podman-compose creates a pod, which means all
>> containers in this pod share the IP 'localhost'. Im my case this was
>> a
>> problem, because the application on one container was looking for
>> resources on the host 'mysql'.
>>
>>
>>
>> _______________________________________________
>> Podman mailing list -- podman(a)lists.podman.io
>> To unsubscribe send an email to podman-leave(a)lists.podman.io
> _______________________________________________
> Podman mailing list -- podman(a)lists.podman.io
> To unsubscribe send an email to podman-leave(a)lists.podman.io
>
_______________________________________________
Podman mailing list -- podman(a)lists.podman.io
To unsubscribe send an email to podman-leave(a)lists.podman.io
9:23 a.m.
Uwe,
I do think my response is applicable. I'm sorry you feel differently.
To use podman generate kube and podman play kube, you don't need to
know any kubernetes. Podman generate simply takes your container or
pod and "records" it in kube yaml; and podman play simply replays it
the way it was recorded. You don't need to have any kubernetes
knowledge.
That said, you are correct that it is a different workflow and does not
cover all container and pod abilities. We are always trying to close
the gap there.
I wish you the best of luck!
Brent
On Mon, 2019-09-02 at 10:09 +0200, Uwe Reh wrote:
Hi Brent,
do you really think, that your article is an appropriate answer to
the
initial question? I suppose not.
Surely you've mad a good job for people which are familiar with
Kubernets. For beginners like Robert and me, it's some kind of
chinese.
Just try to imaging:
- you are relatively new in the area of containers.
- you have heard just enough about Kubernets, to realize that it's
out
of scope for your needs.
- you have read about Kubernet's 'kompose' and have seen your
article.
A beginner like me, may think, "Ok, sounds reasonable".
But this leads into an dead end. 'kompose' creates a lot of nice
detailed configuration files, but none of them seems to be suitable
for
podman...
Without any experience or deeper knowledge about Kubernet, I
resigned.
Uwe
PS.
Maybe, you could write a detailed howto for the new podman pages.
Am 30.08.19 um 13:41 schrieb Brent Baude:
> Depending on the complexity of the task, podman generate kube and
> podman play kube may also be an answer.
>
> https://developers.redhat.com/blog/2019/01/29/podman-kubernetes-yaml/
>
> While most of the article talks about replaying on a kube-basaed
> runtime, one of the last paragraphs talks about replaying locally.
>
> On Fri, 2019-08-30 at 13:09 +0200, Uwe Reh wrote:
> > Hi Robert,
> >
> > I'm not an expert at all. But I had this question a few days ago.
> >
> > The most common tool seems to be
> > https://github.com/containers/podman-compose.
> >
> > Unfortunately, in my case it was not able to replace docker's
> > compose
> > completely¹.
> > But podman-compose has a option to trace what it is trying to do.
> > This
> > helped me, to write a simple script, which is doing the same job
> > as
> > the
> > initial composer file.
> >
> > Uwe
> >
> >
> > ¹) Sorry I haven't documented the problem well.
> > As far as I remember, the problem was that composer creates/runs
> > standalone containers with different IPs and an implicit DNS
> > service
> > (container name -> ip). podman-compose creates a pod, which means
> > all
> > containers in this pod share the IP 'localhost'. Im my case this
> > was
> > a
> > problem, because the application on one container was looking for
> > resources on the host 'mysql'.
> >
> >
> >
> > _______________________________________________
> > Podman mailing list -- podman(a)lists.podman.io
> > To unsubscribe send an email to podman-leave(a)lists.podman.io
> _______________________________________________
> Podman mailing list -- podman(a)lists.podman.io
> To unsubscribe send an email to podman-leave(a)lists.podman.io
>
_______________________________________________
Podman mailing list -- podman(a)lists.podman.io
To unsubscribe send an email to podman-leave(a)lists.podman.io
10:03 a.m.
Hi Brent,
Sorry if my provocation wasn't that small, I intended to.
Podman is for a newbie like English for a none native speaker. It is
easy to do first steps, but until you are really experienced there are
lots of traps to step in.
The point I wanted to express was, "It is nice that podman plays
kube.yaml. But how to get the file to play?"
- There seem to bee no automatic way to translate ONE composer.yaml into
ONE kube.yaml. 'Kubernetes kompose' generates multiple and quire verbose
files.
- If someone is able to write his own kube.yaml, he will have probably
no more need for help in converting a composer file.
Again, it was never my intention to criticize your work. But sometimes
even 'good men' do forget that others may not have the same full
featured background of knowledge.
Uwe
Am 09.09.19 um 16:23 schrieb Brent Baude:
> Uwe,
>
> I do think my response is applicable. I'm sorry you feel differently.
> To use podman generate kube and podman play kube, you don't need to
> know any kubernetes. Podman generate simply takes your container or
> pod and "records" it in kube yaml; and podman play simply replays it
> the way it was recorded. You don't need to have any kubernetes
> knowledge.
>
> That said, you are correct that it is a different workflow and does not
> cover all container and pod abilities. We are always trying to close
> the gap there.
>
> I wish you the best of luck!
10:12 a.m.
Uwe,
Thanks for the feedback. We have definitely been looking at and
thinking about how to ingest Compose yaml. The biggest challenge is
standardization. There's no open standard for Compose. Even though it's
easy to use, it's basically a trap in that it leads you down a path for
multi-container orchestration that is only used by Docker Inc. Everybody
else is standardizing on Kube yaml, and that appears to be the direction of
the future (even though people are still adopting Compose because it's
easy). Basically, we have a race condition where some people are adopting
Compose, or already have Compose yaml, and the rest of the world is moving
to Kube yaml. Compose yaml is easier, Kubernetes is more powerful.
It's a really, really tough situation and one of our #1 requests from the
community. We took a deep look, talked about it a LOT, and decided that
"podman generate kube" and "podman play kube" was the closest we could
come
to giving users:
1. something easy
2. Something compatible with where the world is going
3. Bridges the gap getting from "podman run" to "kubectl create -f
kube.yaml" (which admittedly is painful for a lot of people)
4. Follows open standards
5. Doesn't leave people on a dead end path
All that said, we are still looking at the possibility of a "podman play
compose" which would obviously give you an easier path from Compose to
Kube. The problem is, that sounds really, really good, but it's much harder
to do in reality :-) We need a really good Compose library, and a way to
translate that will be sane. Mileage will definitely vary if we ever
implement that solution...
Best Regards
Scott M
On Wed, Sep 11, 2019 at 11:04 AM Uwe Reh <reh(a)hebis.uni-frankfurt.de> wrote:
Hi Brent,
Sorry if my provocation wasn't that small, I intended to.
Podman is for a newbie like English for a none native speaker. It is
easy to do first steps, but until you are really experienced there are
lots of traps to step in.
The point I wanted to express was, "It is nice that podman plays
kube.yaml. But how to get the file to play?"
- There seem to bee no automatic way to translate ONE composer.yaml into
ONE kube.yaml. 'Kubernetes kompose' generates multiple and quire verbose
files.
- If someone is able to write his own kube.yaml, he will have probably
no more need for help in converting a composer file.
Again, it was never my intention to criticize your work. But sometimes
even 'good men' do forget that others may not have the same full
featured background of knowledge.
Uwe
Am 09.09.19 um 16:23 schrieb Brent Baude:
> Uwe,
>
> I do think my response is applicable. I'm sorry you feel differently.
> To use podman generate kube and podman play kube, you don't need to
> know any kubernetes. Podman generate simply takes your container or
> pod and "records" it in kube yaml; and podman play simply replays it
> the way it was recorded. You don't need to have any kubernetes
> knowledge.
>
> That said, you are correct that it is a different workflow and does not
> cover all container and pod abilities. We are always trying to close
> the gap there.
>
> I wish you the best of luck!
_______________________________________________
Podman mailing list -- podman(a)lists.podman.io
To unsubscribe send an email to podman-leave(a)lists.podman.io
--
--
Scott McCarty, RHCA
Product Management - Containers, Red Hat Enterprise Linux & OpenShift
Email: smccarty(a)redhat.com
Phone: 312-660-3535
Cell: 330-807-1043
Web: http://crunchtools.com
Have questions on Red Hat UBI? Check out the official FAQ:
https://red.ht/2yaUcez
1682
days inactive
1694
days old
8 comments
5 participants
participants (5)
-
Brent Baude -
Daniel Walsh -
Robert P. J. Day -
Scott McCarty -
Uwe Reh