Thread: Postgresql Replication High Availability
I am currently looking into high availability solutions for Postgresql.
I am designing system that should not have a single point of failure. I would like to at
run-time Add new hardware and add them to my (cluster ) environment without shutting anything down.
Please advise, Thanks
#1. Is PGCluster production environment ready?
#2. CJDBC.... has anyone used this in production environment with Postgresql?
#3. Any other solutions viable for production environments?
Yahoo! Music Unlimited - Access over 1 million songs. Try it free.
We're on the same situation. --> PGCluster Actualy, we installed and tested PGCluster on UML (User Mode Linux) environment and real distributed stations. There were a lot of critical issues on it! For example, once a user tries to connect to an unexisting database, the load balancer fails and needs to be restarted... Another: if closed by Ctrl+C, the postmaster starts in an infinite loop! We sent these problems to its User List, and we noted that there are other unsolved issues and a lot of patches not officially posted yet... One thing that made us afraid is that the project isn't updated since march of 2005! http://pgcluster.projects.postgresql.org/ Too much risky to our critical mission database, we thought! --> ClusGres This application should be very promisor, if it weren't "vaporware"... http://www.linuxlabs.com/clusgres.html --> C-JDBC / Sequoia It looks like we're gonna stay with this solution, as I explained on a later e-mail. http://c-jdbc.objectweb.org/ http://sequoia.continuent.org/ ------ Christopher wrote: I am currently looking into high availability solutions for Postgresql. I am designing system that should not have a single point of failure. I would like to at run-time Add new hardware and add them to my (cluster ) environment without shutting anything down. Please advise, Thanks #1. Is PGCluster production environment ready? #2. CJDBC.... has anyone used this in production environment with Postgresql? #3. Any other solutions viable for production environments? ------ Regards, Rodrigo Hjort Linux User #348191 (http://counter.li.org)
Hello Rodrigo, Have you (or anyone else on this list) had great success with Sequoia? I've been reading a bit about it. It looks like it might work for us for a high-availability solution. Do you think it's just as good as using p/cluster? We're deploying to Mac OS X Server and it looks like p/cluster only supports Linux on x86. Thanks, ____________________________________________________________________ Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com ClickSpace Interactive Inc. Suite L100, 239 - 10th Ave. SE Calgary, AB T2G 0V9 http://www.clickspace.com On Dec 2, 2005, at 8:32 PM, Rodrigo Hjort wrote: > We're on the same situation. > > --> PGCluster > > Actualy, we installed and tested PGCluster on UML (User Mode Linux) > environment and real distributed stations. There were a lot of > critical issues on it! For example, once a user tries to connect to an > unexisting database, the load balancer fails and needs to be > restarted... Another: if closed by Ctrl+C, the postmaster starts in an > infinite loop! We sent these problems to its User List, and we noted > that there are other unsolved issues and a lot of patches not > officially posted yet... > > One thing that made us afraid is that the project isn't updated since > march of 2005! > http://pgcluster.projects.postgresql.org/ > > Too much risky to our critical mission database, we thought! > > --> ClusGres > > This application should be very promisor, if it weren't "vaporware"... > http://www.linuxlabs.com/clusgres.html > > --> C-JDBC / Sequoia > > It looks like we're gonna stay with this solution, as I explained on a > later e-mail. > > http://c-jdbc.objectweb.org/ > http://sequoia.continuent.org/ > > ------ > > Christopher wrote: > > I am currently looking into high availability solutions for > Postgresql. > > I am designing system that should not have a single point of failure. > I would like to at > run-time Add new hardware and add them to my (cluster ) environment > without shutting anything down. > > Please advise, Thanks > > #1. Is PGCluster production environment ready? > #2. CJDBC.... has anyone used this in production environment with > Postgresql? > #3. Any other solutions viable for production environments? > > ------ > > Regards, > > Rodrigo Hjort > > Linux User #348191 (http://counter.li.org) > > ---------------------------(end of > broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq >
Attachment
Brendam, We started testing Sequoia last week, and it reached our expectatives of High Availability and Load Balance. We still need to verify if the performance is not so degradated due to multiple clusters updating. What is the difference between Sequoia and p/cluster? I mean, the latter is a commercial solution offered by Continuent, the same producer of Sequoia. Our applications are based on JBoss, connecting to a PostgreSQL via JDBC. One of the greatest features we found on Sequoia is the URL style: "jdbc:sequoia://first_node:port,second_node:port/database" When a Sequoia Controller is down (eg: first_node), the JDBC tries another one (ie, second_node), and it occurs transparently to the application! The JDBC Driver handles it all alone! I don't know if with a pure PostgreSQL JDBC Driver and the Slony I + Heartbeat joint solution the application would perform the same behavior... 2005/12/3, Brendan Duddridge <brendan@clickspace.com>: > Hello Rodrigo, > > Have you (or anyone else on this list) had great success with > Sequoia? I've been reading a bit about it. It looks like it might > work for us for a high-availability solution. Do you think it's just > as good as using p/cluster? We're deploying to Mac OS X Server and it > looks like p/cluster only supports Linux on x86. > > Thanks, > > ____________________________________________________________________ > Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com > > ClickSpace Interactive Inc. > Suite L100, 239 - 10th Ave. SE > Calgary, AB T2G 0V9 > > http://www.clickspace.com Regards, Rodrigo Hjort http://www.pr.gov.br
Hi Rodrigo, So where did you install sequoia? On your application servers? On your database servers? Or on dedicated servers running only sequoia? Thanks, ____________________________________________________________________ Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com ClickSpace Interactive Inc. Suite L100, 239 - 10th Ave. SE Calgary, AB T2G 0V9 http://www.clickspace.com On Dec 4, 2005, at 11:41 AM, Rodrigo Hjort wrote: > Brendam, > > We started testing Sequoia last week, and it reached our expectatives > of High Availability and Load Balance. We still need to verify if the > performance is not so degradated due to multiple clusters updating. > > What is the difference between Sequoia and p/cluster? I mean, the > latter is a commercial solution offered by Continuent, the same > producer of Sequoia. > > Our applications are based on JBoss, connecting to a PostgreSQL via > JDBC. One of the greatest features we found on Sequoia is the URL > style: > > "jdbc:sequoia://first_node:port,second_node:port/database" > > When a Sequoia Controller is down (eg: first_node), the JDBC tries > another one (ie, second_node), and it occurs transparently to the > application! The JDBC Driver handles it all alone! I don't know if > with a pure PostgreSQL JDBC Driver and the Slony I + Heartbeat joint > solution the application would perform the same behavior... > > 2005/12/3, Brendan Duddridge <brendan@clickspace.com>: >> Hello Rodrigo, >> >> Have you (or anyone else on this list) had great success with >> Sequoia? I've been reading a bit about it. It looks like it might >> work for us for a high-availability solution. Do you think it's just >> as good as using p/cluster? We're deploying to Mac OS X Server and it >> looks like p/cluster only supports Linux on x86. >> >> Thanks, >> >> ____________________________________________________________________ >> Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com >> >> ClickSpace Interactive Inc. >> Suite L100, 239 - 10th Ave. SE >> Calgary, AB T2G 0V9 >> >> http://www.clickspace.com > > Regards, > > Rodrigo Hjort > http://www.pr.gov.br > > ---------------------------(end of > broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings >
Attachment
The plan is the 3rd option, to achieve Fail Over capabilities. There will be 2 or more servers running only Sequoia Controller, pointing to a group of PostgreSQL backends each one. The application server will use Sequoia JDBC Driver (not PostgreSQL JDBC anymore) where those controllers will be specified. No more modifications are needed to do it! Well, the exception is a string on the setting XML used by Hibernate... 2005/12/4, Brendan Duddridge <brendan@clickspace.com>: > Hi Rodrigo, > > So where did you install sequoia? On your application servers? On > your database servers? Or on dedicated servers running only sequoia? > > Thanks, > > ____________________________________________________________________ > Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com > > ClickSpace Interactive Inc. > Suite L100, 239 - 10th Ave. SE > Calgary, AB T2G 0V9 > > http://www.clickspace.com > Regards, Rodrigo Hjort GTI - Projeto PostgreSQL CELEPAR - Cia de Informática do Paraná http://www.pr.gov.br