The old virtual_ipaddress_excluded has not been deleted after reconfig vlan interface #2383
Comments
|
You are showing that keepalived segfaulted, which would explain why some addresses failed to be removed (keepalived crashed before it removed the addresses). The version of keepalived you are running is quite old (nearly 4 years), and some 800 non-merge commits behind the current version. Can you please try running with v2.2.8 and see if you still experience the problem. If you still experience the problem with v2.2.8, then a stack backtrace produced from the coredump could help us identify the cause of the problem. |
|
thanks for your reply! @pqarmitage maybe related with this one？#1902 (comment) |
|
@CoderSinger The answer to your question about whether removing an interface causes a segfault has been fixed is that I do not know, although I suspect it has. Unfortunately I do not have the time to go through all the commits, and I don't remember all the changes that have been made over the last 4 years or so! This is why I asked you to try v2.2.8 to see if it resolves your issue. If the issue still exists with v2.2.8 then I will research it provide a fix. If you want to see all the relevant commits, you could try executing in a keepalived git tree: |
|
@pqarmitage Thank you for your reply. I will consider upgrading to the Keepalived version.
The reason why I have this question is because I encountered the following issues when using 3pids for reload. Abnormal checker logs: It's like I repeatedly called reload, causing duplicate deletion of service configuration and resulting in an error. At the same time point in the error log, it was observed in the VRRP log that the keepalived process entered the backup state, but in reality, the keepalived process is executing on a single node. |
|
It is |
|
Thank you, i have no other questions. |
|
Hi, I'm sorry to bother you again due to the same issue. @pqarmitage My operations:
I'm running keepalived in docker container, which's image version is Alpine 3.17.6. There is no SEGFAULT this time. Then the old virtual ip address still exists on network interface. |
|
@codesinger Can you please post the keepalived config from both before the reload and after the reload so that I can be absolutely clear about what changes are being made. I will then test this and make sure I understand what is happening, and I hope provide a fix. |
|
@pqarmitage I will post the new and old configurations and logs below. I have noticed that even after restarting with the new configuration, keepalived will still assign the old address to the network port. I am not sure if this is the problem with the "netlink_if_address_filter" function. Network interface infromation: Old configuration: New configuration: I think the key logs are: |
Describe the bug
when delete vlan interface and create a new one, and reconfig the virtual_ipaddress_excluded at the same time in keepalived.conf, the old virtual_ipaddress_excluded has not been deleted, it's still on the new vlan interface with new ip address.
To Reproduce
It is not inevitable, but multiple attempts can lead to it.
old conf
new conf
restart keepalived(kill -15 {3 pids}, and run keepalived executable file)
Following the above steps may result in the following situations: the old virtual_ipaddress_excluded has not been deleted on new vlan interface
87: eno3.996@eno3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether bc:16:95:04:92:20 brd ff:ff:ff:ff:ff:ff
inet6 3efe:ffff:0:f101::13/64 scope global nodad
valid_lft forever preferred_lft forever
inet6 2efe:ffff:0:f101::13/64 scope global
valid_lft forever preferred_lft forever
inet6 fe80::be16:95ff:fe04:9220/64 scope link
valid_lft forever preferred_lft forever
Expected behavior
The old virtual_ipaddress_excluded should be deleted after reconfig keepalived.conf and restart keepalived
Keepalived version
Distro (please complete the following information):
Details of any containerisation or hosted service (e.g. AWS)
docker container
Configuration file:
Notify and track scripts
no
System Log entries
keepalived.log
keepalived_vrrp.log
Did keepalived coredump?
This phenomenon is not inevitable
Additional context
nothing
The text was updated successfully, but these errors were encountered: