On PTX10002 devices, odd number ports may experience port down even if the link flap (down and up) is shorter than the configured "hold-time down" interval, if the interval is less than 3 seconds.
To prevent this problem from occurring, we recommend that you configure odd number ports on PTX10002 with "hold-time down" greater than 3 seconds to help network stability against link flaps shorter than 3-second period.
On PTX10002 devices, odd number ports may experience port down even if the link flap (down and up) is shorter than the configured "hold-time down" interval, if the interval is less than 3 seconds. Even if no specific configuration exists, the default value is "hold-time down 0" second, which will still lead to the same problem for odd number ports.
For example, configure et-0/0/21 with the hold-time down value as 1000:
[edit] lab@dut# show interfaces et-0/0/21 hold-time up 0 down 1000;
Now, test the port loss-of-signal (LOS). The debug log shows the port down with local fault = 1:
[Wed Sep 7 02:38:50.409 LOG: Info] T6E(0/0) link[21 0] Down: phy_link_ok: 0 mac_lf:1 mac_rf:0 asic_flt: 0, intr:1 chged:1 dfe:2, fcalr:1, fcup:0 tx_flush:0
Even if you control the test to restore the signal and clear the local fault within one second, the debug log does not show the local fault clear in time. The debug log continues to report expiry of the "hold-time down 1 second" value:
[Wed Sep 7 02:38:51.490 LOG: Debug] IFD: et-0/0/21, up/down timer fired (1000) [Wed Sep 7 02:38:51.495 LOG: Debug] IFD: et-0/0/21, sent flap msg to RE, Down state
The behavior is confirmed as a hardware limitation. On PTX10002 devices, all odd ports have a Wireless Area Network (WAN) re-timer connected to it. When an odd number port experiences LOS and the signal is restored within a sub-second, the re-timer does not reset the timer soon after the local fault is cleared but takes more time. So the configured "hold-time down" duration expires and the interface is declared as down.
To address this problem, configure the odd-number ports on PTX10002 devices with the "hold-time down" value greater than 3 seconds to help network stability against link flaps that are shorter than the 3-second period:
# set interfaces et-0/0/1 hold-time up <milliseconds> down 3000 # set interfaces et-0/0/3 hold-time up <milliseconds> down 3000 # set interfaces et-0/0/5 hold-time up <milliseconds> down 3000
2023-01-13: Initial version