Symptom

On ASR9K 7.6.x, after ISSU upgrade with 6000 bgp session in idle mode (already prior to starting ISSU), bgp process will keep accumulating memory for hours until it reaches the RLIMIT and crashes to recover. RP/0/RSP1/CPU0:Jun 10 10:02:32.782 PDT: syslog_dev[117]: bgp[1103] PID-15203: Memdbg failure in process bgp (15172): Rlimit exceeded: current usage 7809031421, limit 7809031440, last allocation 32 RP/0/RSP1/CPU0:Jun 10 10:02:33.113 PDT: dumper[67454]: %OS-SYSLOG-6-LOG_INFO : Dumping core /misc/scratch/core/bgp_15172.by.6.20220610-100232.xr-vm_node0_RSP1_CPU0.c6794.core.gz RP/0/RSP1/CPU0:Jun 10 10:03:02.760 PDT: processmgr[51]: bgp(1) (jid 1103) (pid 15172) (fail_count 1) abnormally terminated, restart scheduled RP/0/RSP1/CPU0:Jun 10 10:03:46.225 PDT: bgp[1103]: %ROUTING-BGP-5-ASYNC_IPC_STATUS : default, process instance 1:(A)inst-id 0, Connection Establised RP/0/RSP1/CPU0:Jun 10 10:03:46.465 PDT: bgp[1103]: %ROUTING-BGP-5-ASYNC_IPC_STATUS : default:(A)inst-id 0, Initial Config Done RP/0/RSP1/CPU0:Jun 10 10:03:54.843 PDT: bgp[1103]: %ROUTING-BGP-5-ADJCHANGE : neighbor 150.1.1.4 Up (VRF: default) (AS: 1.1) RP/0/RSP1/CPU0:Jun 10 10:03:54.847 PDT: bgp[1103]: %ROUTING-BGP-5-NSR_STATE_CHANGE : Changed state to Not NSR-Ready

Conditions

Prior to ISSU, thousands of bgp neighbor are not established.

Workaround

shutdown those neighbors prior to ISSU or configure session-mode passive

Further Problem Description