Hello everybody. I state that I'm not a guru, but the behavior of a test stack leaves me perplexed. I'm probably wrong, can anyone help me?
This is the situation: two M4300-8X8F stacked, ring configuration, 10G copper: ports 15 and 16 of both switches were used. Ports 11 of the two units were configured in LAG, ch1, default options. Ports 12 of the two units were configured in LAG, ch2, default options. The configuration has been saved to the stack. Ports 11 (ch1) of the stack were connected to ports 3 and 4 of an M4300-24X, where the LAG ch1 was enabled, again with standard options. Ports 12 (ch2) of the stack were connected to ports 3 and 4 of an M4300-12X12F, where the LAG ch2 was enabled, again with standard options. Basically there are two switches A and B, connected to each other through stack C and the connection is in LAG for both.
Good. Now I connect the laptop to switch A and ping the IP of switch C. I can disconnect one of the pairs of the LAG without much thought. I can also unplug two cables from the LAG, obviously not from the same LAG. This behavior is what I expect. If I restart one of the stack members via CLI or web interface, I don't lose packages. Great.
The problem is when I restart one of the two members of the stack by unplugging the power cable (failure simulation): I lose at least 3 pings. It only happens the first time: i.e. successive reboots (always in this way, i.e. by removing the power cable) of any of the two members cause me to lose at most a ping. If I shut down the stack completely (both switches), the problem comes back on the first restart of either one and then goes away.
Personally, I expected to lose at most one ping, not 3 or more.
The switch configuration is the default for everything else that hasn't been mentioned here. The settings for the LAG are also the default ones. Only the stack and LAG ports have been set. The firmware is 12.0.17.10 for everyone.
