[LU-16450] Cancel outstanding kfilnd transactions if handshake fails

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Assignee: Chris Horn
• Reporter: Chris Horn

Description

If handshake is sent to a peer that is up but does not have kfilnd started then the handshake fails quickly:

00000800:00000200:14.0:1666723382.289200:0:14469:0:(kfilnd_tn.c:621:kfilnd_tn_state_idle()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 TN_EVENT_TX_HELLO event status 0
00000800:00000200:14.0:1666723382.289204:0:14469:0:(kfilnd_tn.c:663:kfilnd_tn_state_idle()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 Using peer 0@kfi(0x0)
00000800:00000200:14.0:1666723382.289215:0:14469:0:(kfilnd_ep.c:420:kfilnd_ep_post_send()) 1@kfi:1 Transaction ID 000000005bef6f4e: Posted send of 38 bytes to peer 0x0: rc=0
00000800:00000200:14.0:1666723382.289219:0:14469:0:(kfilnd_tn.c:285:kfilnd_tn_state_change()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 TN_STATE_IDLE -> TN_STATE_IMM_SEND state change
00000800:00000200:8.0F:1666723382.289294:0:12476:0:(kfilnd_tn.c:866:kfilnd_tn_state_imm_send()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 TN_EVENT_TX_FAIL event status -5
00000800:00000200:8.0:1666723382.289302:0:12476:0:(kfilnd_tn.c:299:kfilnd_tn_status_update()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 0 -> -5 status change
00000800:00000200:8.0:1666723382.289305:0:12476:0:(kfilnd_tn.c:305:kfilnd_tn_status_update()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 0 -> 10 health status change
00000800:00000200:8.0:1666723382.289310:0:12476:0:(kfilnd_tn.c:1381:kfilnd_tn_free()) KFILND_MSG_HELLO_REQ Transaction ID 000000005bef6f4e: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 Transaction freed

But the transaction that precipitated the handshake will continue to wait for the full LND timeout:

...
00000800:00000200:8.0:1666723423.060053:0:12476:0:(kfilnd_tn.c:645:kfilnd_tn_state_idle()) KFILND_MSG_IMMEDIATE Transaction ID 00000000d0315dad: 1@kfi:1 -> 0@kfi(0000000045a72da0):0x0 0@kfi hello response pending
...