[Geoserver-devel] Getting two GeoServer running against the same data dir... not possible in RC3+?

Hi,
I'm wondering how can we run two GeoServers against the same data dir
with RC3+

With past versions there were a few hurdles, one had to set a custom
log file per instance (there is a documented system variable for that),
and disable GWC metastore to avoid conflicts.

Now there is the quota module that is also preventing two GS+GWC
to startup in parallel, the first will lock the sleepycat db and the second
will fail.

I guess that if it was just standalone GWC one could just remove the
quota module from the classpath, but with the GS integration that
module is referenced by the GWC class in the gwc module.

Is there any way out?
So far it seems the only one is to remove GWC altogheter from GeoServer,
which is a pity...

This looks like a serious regression to me, a setup with multiple GS
per machine is common on higher end hardware where the java2d
scalability issues bites more (thinking machines with 16-32 cores).

So far it seems the only way is to get rid of the integrated GWC
altogheter, which is a pity since the itegrated GWC can seed much
faster than a stand alone one (well, unless the layers are dog slow
to start with).

Btw, there was also a thread about this on the user list where people
complained about the regression, so it's not just me being bitten by it.

Cheers
Andrea

--
-------------------------------------------------------
Ing. Andrea Aime
GeoSolutions S.A.S.
Tech lead

Via Poggio alle Viti 1187
55054 Massarosa (LU)
Italy

phone: +39 0584 962313
fax: +39 0584 962313
mob: +39 333 8128928

http://www.geo-solutions.it
http://geo-solutions.blogspot.com/
http://www.youtube.com/user/GeoSolutionsIT
http://www.linkedin.com/in/andreaaime
http://twitter.com/geowolf

-------------------------------------------------------

uh, oh, sigh. so far so good for iterative development.
Though the end goal is to be able to do real clustering this is a big
inconvenient indeed meanwhile.
I guess the easiest way out would be an extra env var to completely
disable diskquota just like you have to do to disable the metastore?

On Wed, 2011-03-30 at 01:47 +0200, Andrea Aime wrote:

Hi,
I'm wondering how can we run two GeoServers against the same data dir
with RC3+

With past versions there were a few hurdles, one had to set a custom
log file per instance (there is a documented system variable for that),
and disable GWC metastore to avoid conflicts.

Now there is the quota module that is also preventing two GS+GWC
to startup in parallel, the first will lock the sleepycat db and the second
will fail.

I guess that if it was just standalone GWC one could just remove the
quota module from the classpath, but with the GS integration that
module is referenced by the GWC class in the gwc module.

Is there any way out?
So far it seems the only one is to remove GWC altogheter from GeoServer,
which is a pity...

This looks like a serious regression to me, a setup with multiple GS
per machine is common on higher end hardware where the java2d
scalability issues bites more (thinking machines with 16-32 cores).

So far it seems the only way is to get rid of the integrated GWC
altogheter, which is a pity since the itegrated GWC can seed much
faster than a stand alone one (well, unless the layers are dog slow
to start with).

Btw, there was also a thread about this on the user list where people
complained about the regression, so it's not just me being bitten by it.

Cheers
Andrea

--
Gabriel Roldan
groldan@anonymised.com
Expert service straight from the developers

Or rather if you don't want the metastore and are already using the
disable metastore key you probably don't want/can't afford diskquota
either, so we could just assume that?

I'm open to suggestions.

Cheers,
Gabriel

On Wed, 2011-03-30 at 00:04 -0300, Gabriel Roldán wrote:

uh, oh, sigh. so far so good for iterative development.
Though the end goal is to be able to do real clustering this is a big
inconvenient indeed meanwhile.
I guess the easiest way out would be an extra env var to completely
disable diskquota just like you have to do to disable the metastore?

On Wed, 2011-03-30 at 01:47 +0200, Andrea Aime wrote:
> Hi,
> I'm wondering how can we run two GeoServers against the same data dir
> with RC3+
>
> With past versions there were a few hurdles, one had to set a custom
> log file per instance (there is a documented system variable for that),
> and disable GWC metastore to avoid conflicts.
>
> Now there is the quota module that is also preventing two GS+GWC
> to startup in parallel, the first will lock the sleepycat db and the second
> will fail.
>
> I guess that if it was just standalone GWC one could just remove the
> quota module from the classpath, but with the GS integration that
> module is referenced by the GWC class in the gwc module.
>
> Is there any way out?
> So far it seems the only one is to remove GWC altogheter from GeoServer,
> which is a pity...
>
> This looks like a serious regression to me, a setup with multiple GS
> per machine is common on higher end hardware where the java2d
> scalability issues bites more (thinking machines with 16-32 cores).
>
> So far it seems the only way is to get rid of the integrated GWC
> altogheter, which is a pity since the itegrated GWC can seed much
> faster than a stand alone one (well, unless the layers are dog slow
> to start with).
>
> Btw, there was also a thread about this on the user list where people
> complained about the regression, so it's not just me being bitten by it.
>
> Cheers
> Andrea
>

--
Gabriel Roldan
groldan@anonymised.com
Expert service straight from the developers

On Wed, Mar 30, 2011 at 5:04 AM, Gabriel Roldán <groldan@anonymised.com> wrote:

uh, oh, sigh. so far so good for iterative development.

Iterative development is good, it's pushing changes in RCs that is
sometime troublesome
in a moment were we need to close up and ship.
However if this did not happen in a RC I might not have noticed it, I stumbled
into it the hard way because we promised to install a 2.1.0 to a customer
that has RC2 and we cannot upgrade them to RC3 now.

This means that, so far, I have to fork RC2 with the latest fixes to
avoid ripping
away GWC from the cluster.

Though the end goal is to be able to do real clustering this is a big
inconvenient indeed meanwhile.
I guess the easiest way out would be an extra env var to completely
disable diskquota just like you have to do to disable the metastore?

That would be good. The flag should just avoid the activation of
the sleepycat db with associated lock.

In the future, in an ideal world, it would be nice to be able to setup
a number of GS+GWC on the same machine, and replicate that
on different machines, all pointing to the same centralized storage,
and be able to have each write tiles to the disk and in the quota/metadata
storages, so that each element in the cluster can leverage the higher
seeding/dynamic caching speed provided by the integration.
Just dreaming.

Cheers
Andrea

--
-------------------------------------------------------
Ing. Andrea Aime
GeoSolutions S.A.S.
Tech lead

Via Poggio alle Viti 1187
55054 Massarosa (LU)
Italy

phone: +39 0584 962313
fax: +39 0584 962313
mob: +39 333 8128928

http://www.geo-solutions.it
http://geo-solutions.blogspot.com/
http://www.youtube.com/user/GeoSolutionsIT
http://www.linkedin.com/in/andreaaime
http://twitter.com/geowolf

-------------------------------------------------------

On Wed, Mar 30, 2011 at 5:31 AM, Gabriel Roldán <groldan@anonymised.com> wrote:

Or rather if you don't want the metastore and are already using the
disable metastore key you probably don't want/can't afford diskquota
either, so we could just assume that?

I'm open to suggestions.

A system variable for each would be good

Cheers
Andrea

--
-------------------------------------------------------
Ing. Andrea Aime
GeoSolutions S.A.S.
Tech lead

Via Poggio alle Viti 1187
55054 Massarosa (LU)
Italy

phone: +39 0584 962313
fax: +39 0584 962313
mob: +39 333 8128928

http://www.geo-solutions.it
http://geo-solutions.blogspot.com/
http://www.youtube.com/user/GeoSolutionsIT
http://www.linkedin.com/in/andreaaime
http://twitter.com/geowolf

-------------------------------------------------------