Evgeniy Berdnikov via Exim-users wrote: > On Wed, Dec 30, 2020 at 02:25:19PM +0700, Victor Sudakov via Exim-users wrote: > > Evgeniy Berdnikov via Exim-users wrote: > > > Yes, I have. And have you looked at packet time stamps? > > > IMHO, 3-5 us time interval between SYN-ACK and RST is > > > too low for userspace responce. > > > > Well, the whole story is happening within one hypervisor (bhyve) so > > network delays can be very low. > > It does not matter: any hypervisor should perform a context switch bitween > kernel process and userspace process, even for container virtulization. > It's havy task. > > > > However, get strace as Jeremy suggested. > > > I expect ECONNREFUSED on connect(). > > > > Is this ktrace informative https://termbin.com/zjsv ? > > I have never seen such a trace before...
It's FreeBSD's native ktrace. FreeBSD also has dtrace but I have no experience therewith. > > As packet is sent, it may be some problem with TCP_FASTOPEN, probably > with its handling in hypervisor and/or external firewall. You must be a wise man Evgeniy! On the exim host, I set sysctl net.inet.tcp.fastopen.client_enable=0 and the problem seemed to go away. Now if you care :-) please explain what it was and what I've lost by disabling fastopen client. > > BTW, this trace may be more meaningful with time stamps, turn them on > if possible. For its academic value, the same ktrace with Unix timestamps: https://termbin.com/iemr -- Victor Sudakov, VAS4-RIPE, VAS47-RIPN 2:5005/49@fidonet http://vas.tomsk.ru/ -- ## List details at https://lists.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://wiki.exim.org/