Hello, I have upgraded to latest stable release 4.0.2. I still have the problem after 1 or 2 days. The only solution for now is to restart each member of the cluster.
It looks like the issues described here: https://issues.apache.org/jira/browse/TS-1375 https://issues.apache.org/jira/browse/TS-1386 I can see issue #1386 is already patched in 4.0.2. Is there anything I can do or provide to help fixing the problem ? Regards, JB On 19/11/2013 11:38, Jean Baptiste Favre wrote: > I forgot an important information: > I'm using TrafficServer 3.2.5 on Debian Wheezy > > Regards, > Jean-Baptiste Favre > > On 19/11/2013 11:36, Jean Baptiste Favre wrote: >> Hello, >> I'm trying to set up a cluster with 2 TrafficServer nodes. >> I use following configuration: >> LOCAL proxy.local.cluster.type INT 1 >> CONFIG proxy.config.cluster.cluster_port INT 8086 >> CONFIG proxy.config.cluster.rsport INT 8088 >> CONFIG proxy.config.cluster.mcport INT 8089 >> CONFIG proxy.config.cluster.mc_group_addr STRING 224.0.1.37 >> CONFIG proxy.config.cluster.mc_ttl INT 1 >> CONFIG proxy.config.cluster.log_bogus_mc_msgs INT 1 >> CONFIG proxy.config.cluster.ethernet_interface STRING eth0 >> >> Everything is fine so far, except that, from time to time, each node >> opens tons of connections with... itself. Then I can see in logs: >> >> [Nov 19 10:37:58.808] Server {0x2b2f0f9a8700} WARNING: too many >> connections, throttling >> [Nov 19 10:52:07.727] Server {0x2b2f0f9a8700} WARNING: too many >> connections, throttling >> [Nov 19 10:52:08.675] Manager {0x7f37b2ddb720} NOTE: >> [Alarms::signalAlarm] Skipping Alarm: 'too many connections, throttling' >> [Nov 19 11:02:33.561] Server {0x2b2f0f9a8700} WARNING: too many >> connections, throttling >> [Nov 19 11:12:51.726] Server {0x2b2f0f9a8700} WARNING: too many >> connections, throttling >> [Nov 19 11:12:54.213] Manager {0x7f37b2ddb720} NOTE: >> [Alarms::signalAlarm] Skipping Alarm: 'too many connections, throttling' >> >> Problem is, on of the 2 nodes opens connection on its own loopback, the >> other on its own public adress. >> Second problem is that traffic server's instances are not used for now. >> >> I'm sure they share exactly the same configuration since it's >> distributed via Chef. >> >> Could HAproxy load-balancer health checks generate so much connections ? >> >> Any advice welcome, >> Regards, >> Jean-Baptiste Favre >> >> >> >> > > > !DSPAM:528b3f2918423027321187! > >
