Re: Flink k8s native support - pod deployments and upgrades

2023-03-04 Thread ramkrishna vasudevan
Thanks Gyula for the update. This helps a lot.

Regards
Ram

On Sat, Mar 4, 2023 at 12:27 PM Gyula Fóra  wrote:

> In general there are no in place upgrades in Flink for either jobs or
> processes . This is not specific to Kubernetes, Yarn or any deployment
> mode.
>
> Gyula
>
> On Sat, 4 Mar 2023 at 06:51, ramkrishna vasudevan  >
> wrote:
>
> > Thanks for the reply. Yes I get that the operator helps out in upgrades
> > also . But when the operator works with flink k8s native clusters, it can
> > help with savepoints and helps in creating a new upgraded cluster, but
> the
> > existing cluster is not generally upgraded due to the design as how
> > clusters are created.
> >
> > Is my understanding correct?
> >
> > Regards
> > Ram
> >
> > On Fri, Mar 3, 2023 at 9:49 PM Thomas Weise  wrote:
> >
> > > Hi,
> > >
> > > The Flink k8s native integration does not handle upgrades. That's what
> > > flink-kubernates-operator was built for. Please check out:
> > >
> > >
> https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/
> > >
> > > Thanks,
> > > Thomas
> > >
> > >
> > > On Fri, Mar 3, 2023 at 2:02 AM ramkrishna vasudevan <
> > > ramvasu.fl...@gmail.com>
> > > wrote:
> > >
> > > > Hi All,
> > > >
> > > > The native implementation of the App mode and session mode - does not
> > > have
> > > > any replica set .
> > > > Instead it just allows the JM to create TM pods on demand.
> > > >
> > > > This is simple and easy in terms of creation of resources, but for an
> > > > upgrade story, how is this managed? Leaving K8s to manage a replica
> set
> > > > based upgrade might be easier right?
> > > >
> > > > Just wanted to understand how upgrades are handled in native K8s
> mode.
> > > >
> > > > Regards
> > > > Ram
> > > >
> > >
> >
>


Re: Flink k8s native support - pod deployments and upgrades

2023-03-03 Thread Gyula Fóra
In general there are no in place upgrades in Flink for either jobs or
processes . This is not specific to Kubernetes, Yarn or any deployment mode.

Gyula

On Sat, 4 Mar 2023 at 06:51, ramkrishna vasudevan 
wrote:

> Thanks for the reply. Yes I get that the operator helps out in upgrades
> also . But when the operator works with flink k8s native clusters, it can
> help with savepoints and helps in creating a new upgraded cluster, but the
> existing cluster is not generally upgraded due to the design as how
> clusters are created.
>
> Is my understanding correct?
>
> Regards
> Ram
>
> On Fri, Mar 3, 2023 at 9:49 PM Thomas Weise  wrote:
>
> > Hi,
> >
> > The Flink k8s native integration does not handle upgrades. That's what
> > flink-kubernates-operator was built for. Please check out:
> >
> > https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/
> >
> > Thanks,
> > Thomas
> >
> >
> > On Fri, Mar 3, 2023 at 2:02 AM ramkrishna vasudevan <
> > ramvasu.fl...@gmail.com>
> > wrote:
> >
> > > Hi All,
> > >
> > > The native implementation of the App mode and session mode - does not
> > have
> > > any replica set .
> > > Instead it just allows the JM to create TM pods on demand.
> > >
> > > This is simple and easy in terms of creation of resources, but for an
> > > upgrade story, how is this managed? Leaving K8s to manage a replica set
> > > based upgrade might be easier right?
> > >
> > > Just wanted to understand how upgrades are handled in native K8s mode.
> > >
> > > Regards
> > > Ram
> > >
> >
>


Re: Flink k8s native support - pod deployments and upgrades

2023-03-03 Thread ramkrishna vasudevan
Thanks for the reply. Yes I get that the operator helps out in upgrades
also . But when the operator works with flink k8s native clusters, it can
help with savepoints and helps in creating a new upgraded cluster, but the
existing cluster is not generally upgraded due to the design as how
clusters are created.

Is my understanding correct?

Regards
Ram

On Fri, Mar 3, 2023 at 9:49 PM Thomas Weise  wrote:

> Hi,
>
> The Flink k8s native integration does not handle upgrades. That's what
> flink-kubernates-operator was built for. Please check out:
>
> https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/
>
> Thanks,
> Thomas
>
>
> On Fri, Mar 3, 2023 at 2:02 AM ramkrishna vasudevan <
> ramvasu.fl...@gmail.com>
> wrote:
>
> > Hi All,
> >
> > The native implementation of the App mode and session mode - does not
> have
> > any replica set .
> > Instead it just allows the JM to create TM pods on demand.
> >
> > This is simple and easy in terms of creation of resources, but for an
> > upgrade story, how is this managed? Leaving K8s to manage a replica set
> > based upgrade might be easier right?
> >
> > Just wanted to understand how upgrades are handled in native K8s mode.
> >
> > Regards
> > Ram
> >
>


Re: Flink k8s native support - pod deployments and upgrades

2023-03-03 Thread Thomas Weise
Hi,

The Flink k8s native integration does not handle upgrades. That's what
flink-kubernates-operator was built for. Please check out:

https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/

Thanks,
Thomas


On Fri, Mar 3, 2023 at 2:02 AM ramkrishna vasudevan 
wrote:

> Hi All,
>
> The native implementation of the App mode and session mode - does not have
> any replica set .
> Instead it just allows the JM to create TM pods on demand.
>
> This is simple and easy in terms of creation of resources, but for an
> upgrade story, how is this managed? Leaving K8s to manage a replica set
> based upgrade might be easier right?
>
> Just wanted to understand how upgrades are handled in native K8s mode.
>
> Regards
> Ram
>


Flink k8s native support - pod deployments and upgrades

2023-03-02 Thread ramkrishna vasudevan
Hi All,

The native implementation of the App mode and session mode - does not have
any replica set .
Instead it just allows the JM to create TM pods on demand.

This is simple and easy in terms of creation of resources, but for an
upgrade story, how is this managed? Leaving K8s to manage a replica set
based upgrade might be easier right?

Just wanted to understand how upgrades are handled in native K8s mode.

Regards
Ram