http://www.continuent.com com_fireboard en-us Thu, 09 Sep 2010 10:42:42 -0400 http://www.continuent.com/components/com_fireboard/template/default/images/english/emoticons/fb_rsspower.gif http://www.continuent.com 88 http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=773&catid=8 Robert, Thanks for your response. I'll keep an eye out for future releases with these features. Johan... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=772&catid=8 Hi Johan, Currently what you describe is not directly supported but we have had requests from other users to be able to control the load balancing percentage across particular servers, so this seems inline with what you are describing. I expect we will add more features in this area over the next couple of releases. That said, another approach that we take in the commercial version is to use virtual IPs to maintain a specific floating IP for a chosen server for reporting. This is based on rules in... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=771&catid=8 Hello, Apologies if this is covered elsewhere, I can't find anything relevant in the documentation. Is it possible to configure the connector to prefer a server closest to the client? I'm thinking of a situation where there is a master server at one site and a slave server at a second site over a WAN link. So I'd like clients at each site to use their local server for reads if it's available and avoid traversing the WAN. Thanks, Johan... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=770&catid=13 Hi Robin! Tungsten ensures zero data loss in the case of a planned failover (done through the cctrl 'switch' command) by shutting down connectivity and pushing a final flush transaction through the system to ensure the slave is up to date. In the case of unplanned failover, it is more complicated, as you may have data that never gets off the master to the slave that is promoted on failure. In this case the data are not exactly 'lost' but more like 'a little hard to find'. You can find out pretty qu... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=769&catid=13 Robert Hodges wrote: Hi Damir, You can definitely use sysbench against Tungsten. Just point it straight at the master database. ... Cheers, Robert Since Tungsten working on JDBC level, when using MySQL C client API to write to master database, will transaction be fully replicated to slave database? If so, how does it work? Will "zero data loss" still be guaranteed when there's no DRBD like HA support? If so, how will it ensure data guarded? Thanks, Robin... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=768&catid=3 ... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=767&catid=3 ... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=766&catid=2 ... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=765&catid=13 Alex wrote: 1. I understand that I could have a tungsten-ized layer in place, that would take care of request dispatching for me? I could magically query this layer, and writes would go to the master, reads to the slave? It's not quite that magical. We have a number of tricks to get reads and writes to split but you need to decide which one is right for your application. You may need to follow conventions for your transactions to get the best behavior. This is documented in the Connector and SQL... http://www.continuent.com/community/lab-projects/tungsten-commons/forum?func=view&id=764&catid=13 Hi Everyone, First - thanks for your time with these questions. My corp currently uses Master-Slave Pairs on a per project basis, and this is an emergent practice that is costing them lots of unnecessary money. The servers aren't used to capacity, and having just recently stepped in, I'm certain they'd do better with a Fan-Out type of approach for read scaling on the slaves. Because of the nature of the information, which needs to be "authoritative" in real-time, circular replication is not an optio...