Hello, We use LVS on Redhat 5.1 (2.6.18-53.el5) on 2 IBMS X3650 servers. Here are the RPM versions: ipvsadm-1.24-8.1.i386.rpm system-config-cluster-1.0.50-1.3.noarch.rpm piranha-0.8.4-7.el5.i386.rpm The two servers are configured as one master / one backup. They are also configured to hold the virtual service RTSP. This service is load balanced with round robin method between the two RTSP gateways (the real servers are also the two X3650 servers). When a RTSP session is set up, LVS selects the RTSP gateway, the session is initialized by accessing database and then Video server. When established, the RTSP session only access the video server for streaming. The monitoring script chosen to monitor the real server is: RTSP access test and database access test. If the database is no more available, the monitoring script returns a bad answer and the service goes down: the process nanny shuts down the IP:PORT session. Because of this shutting down, all the current streamings on this RTSP gateway stop. Here are the logs: Jan 8 17:28:17 FHS1 nanny[10420]: Trouble. Recieved results are not what we expected from (172.20.253.215) Jan 8 17:28:17 FHS1 nanny[10420]: shutting down 172.20.253.215:554 due to connection failure For robustness reason, we need that nanny process prevents new connections when the database goes down. But we need that current RTSP sessions go on, NOT STOP. Is there any way not to shutdown current connections when the monitoring scripts answer not what expected? Dynamically change weight of the real servers to zero for example? Thank you for your answer. Anne -- redhat-list mailing list unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subjecthttps://www.redhat.com/mailman/listinfo/redhat-list
