Hi,
there is a firewall between our Spacewalk-Server and our Oracle-DB, that silently drops connections that have not been used for one hour. When the Spacewalk-Server tries to reuse Database connections after they have been idle for more than 60 minutes, it won't succeed. It takes quite a while until it gives up reusing such connections and just opens up a new one. When a spacewalk-client tries to contact the spacewalk-server it usually takes more than 10 minutes of constant retries until a connection succeeds. That seems to be the time the spacewalk-server needs to get a new, stable connection to the Database. Since the firewall's behavior can't be adjusted, the most obvious solution would be to enable keepalive probes for the database connection. But I did not find a way to do this. First I tried to set "SQLNET.EXPIRE_TIME = 10" in /etc/rhn/tns_admin/osa-dispatcher/sqlnet.ora - which should tell the Oracle Instant Client to use its implemented keepalives procedures. It just doesn't work. Then I tried to set appropriate TCP-Keepalive Values in the Linux Kernel (sysctl -w net.ipv4.tcp_keepalive_time=600). But the sockets opened by spacewalk don't use "SO_KEEPALIVE" - which effectively disables tcp-keepalive. Can someone help me out? We use spacewalk 1.8 (from the yum repo yum.spacewalkproject.org) on RHEL6 with a separate Oracle 11 database. We use the Oracle Instant Client from the same repo to connect to the DB. Best regards, Stefan
_______________________________________________ Spacewalk-list mailing list Spacewalk-list@redhat.com https://www.redhat.com/mailman/listinfo/spacewalk-list
