Re: [ADMIN] Clustering Postgres

We'd be very interested in the performance, especially with the disk
being used as in interconnect (I assume that that is being used to
simulate global shared memory?)

We did have a talk with the folks over at http://www.linuxlabs.com/clustgres

They were honest enough to say that even with the Dolphon interconnects, that writing would be slower than usual. We
havenot made a complete decision on that yet, but are still looking at alternatives. 

It seems that Clustres might be a good solution. They use a SAN, the Dolphin interconnects, and multiple diskless
nodes.In theory I am sure it works, but this is one of the chicken/egg problems where we want to see it in general use
firstbefore committing :-) 

- Ericson




Greg Spiegelberg wrote:

> 2 dual PIII-1.4GHz Linux RedHat 3.0 servers each with onboard AIC7xxx
> and 2 U160 drives (mirrored), 2 GB of ECC memory, 2 onboard Intel Ether
> Pro 100Mbps type NIC's, 1 Qlogic 2310 HBA on a 1GBs SAN using a Candera
> 510 storage cluster with Xyratex SATA on the backend configured with 1+0
> LUN's.  It will be using a private network and a small disk on the SAN
> as the interconnect.
>
> Once I have it running and I've run the necessary tests for myself
> I could run a few tests for anyone else who is interested.  Deadline
> is June 18th when either I scrap it and try a different solution or
> we have to buy in which case I'll need it for our production databases.
> Below is the config I intend to use for PostgreSQL 7.4.2.
>
> Greg
>
>
> PostgreSQL 7.4.2 config & compile
> configure \
>   --prefix=/psfs1/apps/pgsql-7.4.2 \
>   --enable-thread-safety \
>   --enable-largefiles \
>   --enable-odbc \
>   --enable-multibyte \
>   --enable-shared \
>   --enable-syslog \
>   --with-perl \
>   --with-CXX \
>   --with-pam \
>   --with-maxbackends=128 \
>   --with-tcl \
>   --with-readline \
>   --with-zlib
> make -j 2 CPPFLAGS="-O3 -DALLOW_ABSOLUTE_DBPATHS -DLOCK_DEBUG" all
>
>
>
> Ericson Smith wrote:
>
>> Do you mean that you are actually testing this product now? If so,
>> what's the hardware configuration (nodes, storage server,
>> interconnects) that you are using?
>>
>> Warmest regards, Ericson Smith
>> Tracking Specialist/DBA
>> +-----------------------+-----------------------------------------+
>> | http://www.did-it.com | "3 Years ago I was worth $5M on paper!" |
>> | eric@did-it.com       | "Sweet! Can I have fries with that?"    |
>> | 516-255-0500          |                          - Tom Tomorrow |
>> +-----------------------+-----------------------------------------+
>>
>>
>> Greg Spiegelberg wrote:
>>
>>> Robert,
>>>
>>> I am currently evaluating PolyServe Matrix Server which is a clustering
>>> solution including a clustered file system (mounted read-write
>>> everywhere).
>>>
>>> Anything special anyone wants to know?  I'm using PostgreSQL 7.4.2 in
>>> a Linux cluster.
>>>
>>> Greg
>>>
>>>
>>> Robert Treat wrote:
>>>
>>>> On Thu, 2004-05-20 at 14:30, Kris Kiger wrote:
>>>>
>>>>> Has anyone heard of clusgres or, better yet, had any experience
>>>>> with it? I came across it while searching for active clustering
>>>>> projects.
>>>>>
>>>>> http://www.linuxlabs.com/clusgres.html
>>>>>
>>>>> Thanks in advance for the info!
>>>>>
>>>>
>>>>
>>>> I have heard of it but I've not heard of any reports on these lists of
>>>> how well it works. If you (or anyone else) gives it a spin please
>>>> try to
>>>> post an account to the list and/or if you send me the info it could
>>>> probably make its way on to techdocs.
>>>> Robert Treat
>>>
>>>
>>>
>>>
>>>
>>
>> ---------------------------(end of broadcast)---------------------------
>> TIP 4: Don't 'kill -9' the postmaster
>
>
>