Is it possible to specify a specific slave interface in a bonded pair when using ping?

I have a bonded pair of network interfaces in RHEL5, eth0 and eth2. The eth0 interface is periodically flaky, so I would like to direct pings through that specific slave interface to determine its behavior. I tried "ping -I eth0 <destination>" but that fails with a "connect: No such device" message. Same result if I use "ping -I eth2 <destination>". If I use "ping -I bond0 <destination>" it works fine.

I think pinging slave interface is not logical . Once bond interface become then bond driver will communicate to ethernet drivers . Direct communication should not be allowed .

Thanks

@vishesh: I understand you point, but it doesn't address the actual issue. I'm trying to determine a way to check the secondary interface's connection status without downing the primary interface. This means not just the status of the interface itself, but that traffic can be routed via it to an external point. The scenario I'm trying to avoid is that the secondary interface cannot route traffic (for example, if the switch port to which it is connected has failed, but the link is still showing as "UP") for some time, and then the primary interface fails. At which point, I've lost complete connectivity, which basically invalidates one part of the reason for bonding - increased availability of the server.

Was this question ever answered? I have the same question, for a similar reason and I'm not finding an answer anywhere. How can I verify that traffic through one particular slave interface of a bonded pair is able to pass traffic end-to-end to a destination on the other side of the switch?

Thanks!

Unfortunately, I never received an answer to my question, and further research seems to indicate it is not possible.