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
Recent containerlab versions start Linux containers with on-failure restart policy. Docker understands that as "restart after a server crash", restarting FRRouting (or IOSv) containers with no usable connectivity (because containerlab uses handcrafted vEth pairs).
The only sane way to fix this is to change the on-failure restart policy. The next containerlab version will include a configurable option (see srl-labs/containerlab#2238) but I don't want to force everyone to upgrade containerlab together with netlab (although we are testing containerlab version during virtualization provider probes, so we could enforce that).
Another option would be to execute docker commands to change the policy in the clab post_start_lab hook, making this fix containerlab-version independent.
Containerlab sets the Docker restart policy of Linux-based containers
to 'on-failure' which restarts all lab nodes with kind=linux (FRR, IOSv)
after a server reboot without the necessary connectivity (veth pairs)
The containerlab 0.59.0 release added 'restart-policy' keyword. That
keyword is now part of configurable 'clab' parameters and is set by
default to 'no' on Linux containers.
Template fixes include indentation cleanup; documentation fixes
include URLs to various clab parameters
Fixes#1399
Recent containerlab versions start Linux containers with on-failure restart policy. Docker understands that as "restart after a server crash", restarting FRRouting (or IOSv) containers with no usable connectivity (because containerlab uses handcrafted vEth pairs).
The only sane way to fix this is to change the on-failure restart policy. The next containerlab version will include a configurable option (see srl-labs/containerlab#2238) but I don't want to force everyone to upgrade containerlab together with netlab (although we are testing containerlab version during virtualization provider probes, so we could enforce that).
Another option would be to execute docker commands to change the policy in the clab post_start_lab hook, making this fix containerlab-version independent.
Any thoughts @hellt @jbemmel @DanPartelly @ssasso?
The text was updated successfully, but these errors were encountered: