[vpsFree.cz: community-list] Stopping vps takes some time

[Jakub Skokan]

On 12/28/19 1:32 AM, Wolf wrote:
> I have an small issue when stopping Alpine (3.11) vps on staging.
> It currently takes long time (multiple minutes, I assume before some
> timeout happens and it is killed).
> 
> When I have remote console opened when stopping, I see this output:
> 
>   * Stopping sshd ... [ ok ]
>   * Shutting down ssh connections ... * sshd: caught SIGTERM, aborting
>   [ ok ]
>   * Stopping busybox syslog ... [ ok ]
>   * Unmounting cgroups ... [ ok ]
> The system is going down NOW!
> Sent SIGTERM to all processes
> 
> and it is stuck like this. Did anyone else notice this and is there
> some known workaround (it is annoying...)?

For those who also noticed this issue, it was caused by the init system 
calling sync() syscall on shutdown. Even when called from within a VPS 
(which is a container), all filesystems on the node were synchronized, 
i.e. including filesystems in all adjacent VPSes. That's how it's 
implemented in Linux and it's why it was so slow. We've patched it in 
the kernel and deployed it yesterday.

Jakub