fyi That linger change was included in 3.3.4 and 3.4.1. Patrick
On Mon, Dec 19, 2011 at 8:44 AM, Benjamin Reed <[email protected]> wrote: > which version of the server are you using. this sounds like the linger > bug that was fixed a while back. > > ben > > On Sun, Dec 18, 2011 at 6:22 PM, Camille Fournier <[email protected]> wrote: >> That is not what we would expect from the cluster. What do the server logs >> show? >> >> On Sun, Dec 18, 2011 at 9:13 PM, 朱晨杰 <[email protected]> wrote: >>> Hi, everyone: >>> I'm now using Zookeeper as my service server. There are 3 Zookeeper >>> servers, then I use 3 pc connect to them. On each pc there are 20 processes >>> connect to zookeeper cluster so there are totally 60 connection. I >>> terminate 20 processes on one pc simultaneously, then I find the status of >>> all the 60 connection turned to "disconnected". After a while, the living >>> 40 process start re-connect to zookeeper, but they all become "expired". I >>> thought the connetion of one client would not affect others, but the result >>> shows it does affect. >>> So I want to ask: Does it cost a lot for zookeeper to deal with a lost >>> connnetion? And if too much connection lost at the same time will cause >>> zookeeper deny of service? >>> Thanks! >>> -- >>> Zhu Chenjie >>> Zhejiang University, China
