Re: [tipc-discussion] [net-next] tipc: reduce sensitive to retransmit failures

[Jon M. <jon...@er...>]

Acked. But you *must* use the monitor functionality for any cluster > 100. Otherwise this is never going to work.

BR
///jon


> -----Original Message-----
> From: Hoang Le <hoa...@de...>
> Sent: 30-Oct-19 02:26
> To: Jon Maloy <jon...@er...>; ma...@do...; tip...@li...;
> yin...@wi...
> Subject: [net-next] tipc: reduce sensitive to retransmit failures
> 
> With huge cluster (e.g >200nodes), the amount of that flow:
> gap -> retransmit packet -> acked will take time in case of STATE_MSG
> dropped/delayed because a lot of traffic. This lead to 1.5 sec tolerance
> value criteria made link easy failure around 2nd, 3rd of failed
> retransmission attempts.
> 
> Instead of re-introduced criteria of 99 failed retransmissions to fix the
> issue, we increase failure detection timer to ten times tolerance value.
> 
> Fixes: 77cf8edbc0e7 ("tipc: simplify stale link failure criteria")
> Signed-off-by: Hoang Le <hoa...@de...>
> ---
>  net/tipc/link.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/net/tipc/link.c b/net/tipc/link.c
> index 7d7a66178607..9f524c325c0d 100644
> --- a/net/tipc/link.c
> +++ b/net/tipc/link.c
> @@ -1084,7 +1084,7 @@ static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
>  		return false;
> 
>  	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
> -			msecs_to_jiffies(r->tolerance)))
> +			msecs_to_jiffies(r->tolerance * 10)))
>  		return false;
> 
>  	hdr = buf_msg(skb);
> --
> 2.20.1