Customer reported that sometimes MX304 started suddenly flapping LAG interfaces while the ae bundle member links are up.

From logs we can see before ae13 LACPD_TIMEOUT, there were nothing special in the log and customer stated everytimes these are happening with BGP flowspec updates were applied.

Feb 8 17:38:40 mx304-fpc0 mgd[13331]: UI_LOGOUT_EVENT: User 'root' logout

Feb 8 18:09:50 mx304 lacpd[21190]: LACPD_TIMEOUT: et-0/0/4: lacp current while timer expired current Receive State: CURRENT

Feb 8 18:09:50 mx304 lacpd[21190]: LACP_INTF_DOWN: ae13: Interface marked down due to lacp timeout on member et-0/0/4

Feb 8 18:09:50 mx304 kernel: _lag_bundlestate_ifd_change: bundle ae13: bundle IFD minimum bandwidth or minimum links not met, Bandwidth

Feb 8 18:09:52 mx304 fpc0 cmerror_set_error: Module: MQSS(1), pfe id:1, error_id:0x228873 reason:MQSS(1): WI-0: OGE 0 interface FIFO overflow 

Feb 8 18:09:52 mx304-fpc0 resiliencyd[10341]: CMError: /fpc/0/platformd/0/cm/0/mqss/1/YTCHIP_MQSS_CMERROR_WI_INT_REG_OGE0_IF_FIFO_OVF (0x228873), scope: pfe, category: functional, severity: major, module: MQSS(1), type: WI: OGE 0 interface FIFO overflow, oc_category: default

Feb 8 18:09:52 mx304-fpc0 resiliencyd[10341]: Performing action get-state for error /fpc/0/platformd/0/cm/0/mqss/1/YTCHIP_MQSS_CMERROR_WI_INT_REG_OGE0_IF_FIFO_OVF (0x228873) in module: MQSS(1) with scope: pfe category: functional level: major, oc_category: default

Feb 8 18:09:54 mx304 fpc0 cmerror_set_error: Module: MQSS(1), pfe id:1, error_id:0x228873 reason:MQSS(1): WI-0: OGE 0 interface FIFO overflow 

Customer stated they are using BGP VPN flowspec for DDOS protection and when ae bundle flapped there were BGP flowspec updates.

In the lab we reproduced similar LACPD timeout issue and it is clearly the case of over subscription of BGP flow spec terms which has lot of flag matches which makes DMEM filter works like serially. 

Fast lookup filters can boost filtering performance by as much as three to four times for filters under 3000 terms.

After we enabled BGP flowspec with FLT(fast-lookup-filter) feature then PPE utilization much lower and no more LACP timeout is observed.

Customer enabled fast-lookup-filter feature for the BGP VPN Flowspec and the issue is not noticed anymore.

Syntax:

set routing-options flow fast-lookup-filter

For VRF:

set routing-instances <vrf-name> routing-options flow fast-lookup-filter