jeffdennison
Technical User
I have recently upgraded a soft real time system from AIX 4.1.5 and am experiencing socket errors in a custom application that were not present in 4.1.5. The system consists of 2 C20 RS/6000s with AIX on a FDDI backbone
with 2 CISCO routers. The C20s act as custom gateways to a set of IBM G3 mainframes running a custom 370 channel protocol. The CISCO routers run OSPF and export default routes to AIX in each C20 via OSPF configed in gated.
The routers are are redundant in that each router is connected to both the FDDI backbone with the C20s as well as to the ethernet subnets where the "clients" reside. The "clients" connect to the C20's via a TCP/IP API.
The problem I'm having is that when I fail the CISCO router that is currently the default gateway all the TCP/IP connections from the C20s to clients go down ("no route to network" errors on sockets in the C20s); eventually (about 4-6seconds) the new default route is imported from the remaining router and the clients can reconnect but this wreaks havok in a real time safety critical system.
In 4.1.5 no errors occurred. Although I'm working on upgrading the application to tolerate errors more robustly, I didn't expect such a big difference betweeen OS versions.
Anyone have any suggestions??? Any help would be greatly appreciated.
Jeff Dennison [sig][/sig]
with 2 CISCO routers. The C20s act as custom gateways to a set of IBM G3 mainframes running a custom 370 channel protocol. The CISCO routers run OSPF and export default routes to AIX in each C20 via OSPF configed in gated.
The routers are are redundant in that each router is connected to both the FDDI backbone with the C20s as well as to the ethernet subnets where the "clients" reside. The "clients" connect to the C20's via a TCP/IP API.
The problem I'm having is that when I fail the CISCO router that is currently the default gateway all the TCP/IP connections from the C20s to clients go down ("no route to network" errors on sockets in the C20s); eventually (about 4-6seconds) the new default route is imported from the remaining router and the clients can reconnect but this wreaks havok in a real time safety critical system.
In 4.1.5 no errors occurred. Although I'm working on upgrading the application to tolerate errors more robustly, I didn't expect such a big difference betweeen OS versions.
Anyone have any suggestions??? Any help would be greatly appreciated.
Jeff Dennison [sig][/sig]