Ldpd - 100% cpu utilisation - How do I restart FRR services?

Hi,

I’ve opened the issue with FRRouting regrading ldpd running at 100% cpu utilisation:

How do I restart just this frr component, instead of restarting the whole VyOS instance?

I see /usr/lib/frr/frrinit.sh launches watchfrr, do I simply kill the ldpd process and hope it gets respawned?

Regards
David Herselman

Hi. In the latest rolling release
restart frr

Many thanks, is there a way to avoid restarting bgpd?

You can restart some components:

sever@vyos01:~$ restart frr 
Possible completions:
  <Enter>       Execute the current command
  bfdd          Restart Bidirectional Forwarding Detection daemon
  bgpd          Restart Border Gateway Protocol daemon
  ospf6d        Restart OSPFv3 daemon
  ospfd         Restart OSPFv2 daemon
  ripd          Restart Routing Information Protocol daemon
  ripngd        Restart RIPng daemon
  staticd       Restart Static Route daemon
  zebra         Restart IP routing manager daemon

Thanks, the ldpd process unfortunately isn’t part of this list though…

Okay, so killing a single daemon results in watchfrr recovering it automatically. It also shows the command one can use to restart the service.

/usr/lib/frr/watchfrr.sh restart ldpd

Logs when simply using ‘kill -9 962’ to kill the process:

Jan 28 16:29:45 testing zebra[930]: [EC 4043309121] Client 'ldp' encountered an error and is shutting down.
Jan 28 16:29:45 testing watchfrr[896]: [EC 268435457] ldpd state -> down : read returned EOF
Jan 28 16:29:45 testing zebra[930]: client 57 disconnected. 0 ldp routes removed from the rib
Jan 28 16:29:45 testing zebra[930]: client 62 disconnected. 0 ldp routes removed from the rib
Jan 28 16:29:50 testing watchfrr[896]: [EC 100663303] Forked background command [pid 12885]: /usr/lib/frr/watchfrr.sh restart ldpd
Jan 28 16:29:50 testing zebra[930]: client 57 says hello and bids fair to announce only ldp routes vrf=0
Jan 28 16:29:50 testing watchfrr[896]: ldpd state -> up : connect succeeded
Jan 28 16:29:51 testing zebra[930]: Extended Error: Label >= configured maximum in platform_labels
Jan 28 16:29:51 testing zebra[930]: [EC 4043309093] netlink-dp (NS 0) error: Invalid argument, type=RTM_NEWROUTE(24), seq=13496381, pid=2982329979
Jan 28 16:29:51 testing zebra[930]: [EC 4043309108] LSP Install Failure: in-label 4294967295

Regards
David Herselman

We have not implemented ldpd in our rolling releases by default, so there are no such additional commands (for restarting).

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.