You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If a router node is not online when the tailscale client logs in, the client does not route the traffic to that router node in case of failover. In that situation, when the node that is advertising routes goes offline, its advertised routes are still flagged as primary.
Expected Behavior
If the primary subnet router node disappear, the failover should kick-in immediately, even if the failover node came online after the client authenticated with headscale.
Steps To Reproduce
Topology
We have one headscale server and 2 router nodes both advertising the same routes.
Our router nodes are subnets routers, we do not route internet traffic through them.
We have a tailscale client running on MacOS that need to access an IP routed by the route nodes.
Case 1 : router node comes online after tailscale client
From a tailscale client (macOS, 1.76.1 standalone variant) there is a ping running to a routed IP : 192.168.168.1
We stop the primary : router-node-i-01cd8b42c5599852d
The ping only misses one beat :
64 bytes from 192.168.168.1: icmp_seq=1708 ttl=64 time=26.200 ms
64 bytes from 192.168.168.1: icmp_seq=1709 ttl=64 time=26.138 ms
64 bytes from 192.168.168.1: icmp_seq=1710 ttl=64 time=29.193 ms
Request timeout for icmp_seq 1711
64 bytes from 192.168.168.1: icmp_seq=1712 ttl=64 time=94.161 ms
Is this a support request?
Is there an existing issue for this?
Current Behavior
If a router node is not online when the tailscale client logs in, the client does not route the traffic to that router node in case of failover. In that situation, when the node that is advertising routes goes offline, its advertised routes are still flagged as primary.
Expected Behavior
If the primary subnet router node disappear, the failover should kick-in immediately, even if the failover node came online after the client authenticated with headscale.
Steps To Reproduce
Topology
We have one headscale server and 2 router nodes both advertising the same routes.
Our router nodes are subnets routers, we do not route internet traffic through them.
We have a tailscale client running on MacOS that need to access an IP routed by the route nodes.
Case 1 : router node comes online after tailscale client
From a tailscale client (macOS, 1.76.1 standalone variant) there is a ping running to a routed IP : 192.168.168.1
Nodes:
Routes:
We start a 2nd router node (router-node-i-065a1cf0840b26f0d)
Nodes
Routes
From the macos tailscale client we see the 2 router-nodes
We stop
router-node-i-0380d27217348885b
(which has the primary route)On the MacOs tailscale client the ping stop working
Nodes
Routes
router-node-i-01cd8b42c5599852d is still advertising as primary
The macos client still show 2 node
We disconnect and reconnect the macos client and the ping comes back
Nodes
Routes
What if the two nodes are up when the MacOS tailscale clients logs in
Ping is continuously running on the MacOS tailscale client
Nodes
Routes
We stop the primary :
router-node-i-01cd8b42c5599852d
The ping only misses one beat :
Nodes
Routes
The route fails over correctly (primary is now
router-node-i-0380d27217348885b
) !re-starting
router-node-i-01cd8b42c5599852d
Nodes
(no effect on the ping)
Routes
stop both router-nodes
Nodes
As soon as we start
router-node-i-01cd8b42c5599852d
ping recover as expected.
Environment
Runtime environment
Anything else?
Other details for the setup
ACLs used :
router-node tailscale parameters :
The text was updated successfully, but these errors were encountered: