You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lenya.apache.org by Michael Wechner <mi...@wyona.com> on 2005/06/23 17:16:13 UTC
Cluster and Flow
Hi
I have learned yesterday that flow is not serializable, which means
it cannot be used within a clustered environment.
I think having Lenya working within a cluster is very important for Lenya.
One way is of course to make flow serializable and the other is not to
use flow.
Anyway, I think it's good to keep this situation in mind.
Thanks
Michi
--
Michael Wechner
Wyona - Open Source Content Management - Apache Lenya
http://www.wyona.com http://lenya.apache.org
michael.wechner@wyona.com michi@apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Michael Wechner <mi...@wyona.com>.
Torsten Schlabach wrote:
>>it seems quite a few people are unaware that there is a significant
>>difference between load balancing and clustering.
>>
>>
>
>Will load-balancing work for authoring?
>
Hardly, I rather doubt it.
First, it depends if the load-balancer is sticky or not.
Second also the locks re writing needs to be centralized and content itself.
Third, I am sure there are more issues.
> Will it work for anything which
>depends on a user session such as continuations, for example?
>
>
No, because sessions are not being replicated. Let's say you order something
and the server goes down, your session will be gone and you will be very
unhappy ;-)
The more Lenya goes into the "Applications Area" the more necessary
clustering
becomes. And Lenya needs to head into this direction to stay competitive.
Michi
>Regards,
>Torsten
>
>
>
>>>>>it's not going to be rare, but rather essential in the near future
>>>>>
>>>>>
>>it seems quite a few people are unaware that there is a significant
>>difference between load balancing and clustering. while i agree that
>>load balancing is a good thing to have (and can already be done, for
>>instance with the techniques outlined in
>>http://wiki.apache.org/cocoon/LoadBalancingWithModProxy), the need for
>>clustering is far less certain.
>>
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
>>For additional commands, e-mail: dev-help@lenya.apache.org
>>
>>
>>
>>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
>For additional commands, e-mail: dev-help@lenya.apache.org
>
>
>
>
--
Michael Wechner
Wyona - Open Source Content Management - Apache Lenya
http://www.wyona.com http://lenya.apache.org
michael.wechner@wyona.com michi@apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Michael Wechner <mi...@wyona.com>.
Gregor J. Rothfuss wrote:
> session replication (and clustering) is *not* something that lenya
> needs in the near future (if ever).
I strongly disagree. Please see my arguments within the anser to Torsten
Michi
--
Michael Wechner
Wyona - Open Source Content Management - Apache Lenya
http://www.wyona.com http://lenya.apache.org
michael.wechner@wyona.com michi@apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by "Gregor J. Rothfuss" <gr...@apache.org>.
Torsten Schlabach wrote:
>> it seems quite a few people are unaware that there is a significant
>> difference between load balancing and clustering.
>
> Will load-balancing work for authoring? Will it work for anything which
> depends on a user session such as continuations, for example?
yes, because sessions do not need to be replicated, they are sticky
(they always go to the same instance in the load-balanced setup). you
really don't want to have to deal with session replication if you can
avoid it. session replication (and clustering) is *not* something that
lenya needs in the near future (if ever). clustering is about as useful
as deploying a full j2ee container just for servlets, ie not useful at
all. funnily, the perceived need for the two often comes together..
that said, jcr should definitely be integrated to protect the sitetree
from concurrent modification. i'd advise to spend time worrying about
that instead of session replication.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Torsten Schlabach <ts...@apache.org>.
> it seems quite a few people are unaware that there is a significant
> difference between load balancing and clustering.
Will load-balancing work for authoring? Will it work for anything which
depends on a user session such as continuations, for example?
Regards,
Torsten
>>>> it's not going to be rare, but rather essential in the near future
>
> it seems quite a few people are unaware that there is a significant
> difference between load balancing and clustering. while i agree that
> load balancing is a good thing to have (and can already be done, for
> instance with the techniques outlined in
> http://wiki.apache.org/cocoon/LoadBalancingWithModProxy), the need for
> clustering is far less certain.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
> For additional commands, e-mail: dev-help@lenya.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by "Gregor J. Rothfuss" <gr...@apache.org>.
>>> it's not going to be rare, but rather essential in the near future
it seems quite a few people are unaware that there is a significant
difference between load balancing and clustering. while i agree that
load balancing is a good thing to have (and can already be done, for
instance with the techniques outlined in
http://wiki.apache.org/cocoon/LoadBalancingWithModProxy), the need for
clustering is far less certain.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Andreas Hartmann <an...@apache.org>.
Torsten Schlabach wrote:
[...]
>>Even if technical people agree that
>>failover on an authoring server is not essential, management people
>>will generally not.
>
>
> Well, if you have a larger community of editors (think of one of the major
> portals for example) then availability of the authoring server can become
> quite important as you will loose a lot of productivity if your authors
> cannot work and - even worse - you might not be able to produce content on
> time.
Actually I meant auto failover without losing the session. Availability
raises another question (related to the repository), but IMO losing the
session when an authoring server dies should be acceptable for an editor.
An end user should certainly not be logged out when a machine dies, so
the session has to be transferred to another server.
-- Andreas
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Torsten Schlabach <ts...@apache.org>.
>> it's not going to be rare, but rather essential in the near future
>
> +1
+1
> Looking at http://lenya.apache.org/roadmap.html, industrial strength
> is one of our "core competences".
This is my reasoning for the +1. I often spend time explaining to people
why Lenya is superiour to PHP / MySQL based CRM systems.
> Even if technical people agree that
> failover on an authoring server is not essential, management people
> will generally not.
Well, if you have a larger community of editors (think of one of the major
portals for example) then availability of the authoring server can become
quite important as you will loose a lot of productivity if your authors
cannot work and - even worse - you might not be able to produce content on
time.
Regards,
Torsten
> Michael Wechner wrote:
>> Gregor J. Rothfuss wrote:
>>
>>> Michael Wechner wrote:
>>>
>>>>> not using flow is of course not an option.
>>>>
>>>>
>>>>
>>>> says who? I think it's of great importance that the community is
>>>> aware of the situation and has the capability of discussing things
>>>> before going ahead and
>>>> jumping over the cliff, which might sound a bit dramatic ;-)
>>>
>>> once you take the time to understand the 1.4 code base, you'll see
>>> why. the (rare)
>>
>> it's not going to be rare, but rather essential in the near future
>
> +1
>
> Looking at http://lenya.apache.org/roadmap.html, industrial strength
> is one of our "core competences". Even if technical people agree that
> failover on an authoring server is not essential, management people
> will generally not. And if we want to use flow for live applications,
> it's quite essential anyway.
>
>>> case of clustering does *not* mean we can inflict pain on everyone
>>> else.
>>
>> I agree re the pain, and it might was a bit provocative, but we all need
>> to
>> be aware of these issues
>>
>>> if someone needs clustering then they should be prepared to do the
>>> necessary changes.
>>
>> it's important that we get started somewhere and not introduce more
>> stuff
>> which doesn't comply to clustering
>
> Yes, we have to be careful. But rather than looking for alternative
> solutions, I'd invest in solving the issue in Cocoon. From my perception
> the pressure on the Cocoon community re. this problem is increasing.
>
> Just my $0.02, but I'd love to see Lenya ready for clustering.
>
> -- Andreas
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
> For additional commands, e-mail: dev-help@lenya.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Andreas Hartmann <an...@apache.org>.
Michael Wechner wrote:
> Gregor J. Rothfuss wrote:
>
>> Michael Wechner wrote:
>>
>>>> not using flow is of course not an option.
>>>
>>>
>>>
>>> says who? I think it's of great importance that the community is
>>> aware of the situation and has the capability of discussing things
>>> before going ahead and
>>> jumping over the cliff, which might sound a bit dramatic ;-)
>>
>> once you take the time to understand the 1.4 code base, you'll see
>> why. the (rare)
>
> it's not going to be rare, but rather essential in the near future
+1
Looking at http://lenya.apache.org/roadmap.html, industrial strength
is one of our "core competences". Even if technical people agree that
failover on an authoring server is not essential, management people
will generally not. And if we want to use flow for live applications,
it's quite essential anyway.
>> case of clustering does *not* mean we can inflict pain on everyone else.
>
> I agree re the pain, and it might was a bit provocative, but we all need to
> be aware of these issues
>
>> if someone needs clustering then they should be prepared to do the
>> necessary changes.
>
> it's important that we get started somewhere and not introduce more stuff
> which doesn't comply to clustering
Yes, we have to be careful. But rather than looking for alternative
solutions, I'd invest in solving the issue in Cocoon. From my perception
the pressure on the Cocoon community re. this problem is increasing.
Just my $0.02, but I'd love to see Lenya ready for clustering.
-- Andreas
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Michael Wechner <mi...@wyona.com>.
Gregor J. Rothfuss wrote:
> Michael Wechner wrote:
>
>>> not using flow is of course not an option.
>>
>>
>> says who? I think it's of great importance that the community is
>> aware of the situation and has the capability of discussing things
>> before going ahead and
>> jumping over the cliff, which might sound a bit dramatic ;-)
>
>
> once you take the time to understand the 1.4 code base, you'll see
> why. the (rare)
it's not going to be rare, but rather essential in the near future
> case of clustering does *not* mean we can inflict pain on everyone else.
I agree re the pain, and it might was a bit provocative, but we all need to
be aware of these issues
> if someone needs clustering then they should be prepared to do the
> necessary changes.
it's important that we get started somewhere and not introduce more stuff
which doesn't comply to clustering
> (which takes a lot more than just making flow serializable, for a
> start read up here):
>
> http://del.icio.us/gregorrothfuss/performance
it doesn't seem to respond ...
Michi
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
> For additional commands, e-mail: dev-help@lenya.apache.org
>
>
--
Michael Wechner
Wyona - Open Source Content Management - Apache Lenya
http://www.wyona.com http://lenya.apache.org
michael.wechner@wyona.com michi@apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by "Gregor J. Rothfuss" <gr...@apache.org>.
Michael Wechner wrote:
>> not using flow is of course not an option.
>
> says who? I think it's of great importance that the community is aware
> of the situation and has the capability of discussing things before
> going ahead and
> jumping over the cliff, which might sound a bit dramatic ;-)
once you take the time to understand the 1.4 code base, you'll see why.
the (rare) case of clustering does *not* mean we can inflict pain on
everyone else. if someone needs clustering then they should be prepared
to do the necessary changes. (which takes a lot more than just making
flow serializable, for a start read up here):
http://del.icio.us/gregorrothfuss/performance
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Michael Wechner <mi...@wyona.com>.
Gregor J. Rothfuss wrote:
> Michael Wechner wrote:
>
>> I have learned yesterday that flow is not serializable, which means
>> it cannot be used within a clustered environment.
>>
>> I think having Lenya working within a cluster is very important for
>> Lenya.
>> One way is of course to make flow serializable and the other is not
>> to use flow.
>
>
> not using flow is of course not an option.
says who? I think it's of great importance that the community is aware
of the situation and has the capability of discussing things before
going ahead and
jumping over the cliff, which might sound a bit dramatic ;-)
> the cocooners have been aware of this for a while, and its one of the
> SOC proposals.
>
> http://wiki.apache.org/cocoon/FlowscriptAndSessionReplication
that's a good start, but we don't know yet if the proposal is going to be
accepted and even if it is, if it will be implemented successfully ...
Michi
--
Michael Wechner
Wyona - Open Source Content Management - Apache Lenya
http://www.wyona.com http://lenya.apache.org
michael.wechner@wyona.com michi@apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by "Gregor J. Rothfuss" <gr...@apache.org>.
Michael Wechner wrote:
> I have learned yesterday that flow is not serializable, which means
> it cannot be used within a clustered environment.
>
> I think having Lenya working within a cluster is very important for Lenya.
> One way is of course to make flow serializable and the other is not to
> use flow.
not using flow is of course not an option. the cocooners have been aware
of this for a while, and its one of the SOC proposals.
http://wiki.apache.org/cocoon/FlowscriptAndSessionReplication
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org
Re: Cluster and Flow
Posted by Andreas Hartmann <an...@apache.org>.
Michael Wechner wrote:
> Hi
>
> I have learned yesterday that flow is not serializable, which means
> it cannot be used within a clustered environment.
>
> I think having Lenya working within a cluster is very important for Lenya.
> One way is of course to make flow serializable and the other is not to
> use flow.
Another way: Provide a backend system for flow context storage.
I don't know if that can be implemented at application level.
-- Andreas
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org