Date: Wed, 30 Dec 2020 16:57:48 +0700 From: Eugene Grosbein <eugen@grosbein.net> To: Victor Gamov <vit@otcnet.ru>, freebsd-net@freebsd.org Subject: Re: 'dropped due to full socket buffers' by SNMP Message-ID: <6c780827-e764-8053-356b-a921e0892c15@grosbein.net> In-Reply-To: <a69d872a-f21e-de66-7677-58beccb0f023@otcnet.ru> References: <388da9a7-7b89-89b2-54eb-17d0e818c924@otcnet.ru> <4e41c1d2-19bc-0345-0b03-526e4cb785c7@otcnet.ru> <b279e676-c789-2978-98a0-b8a4b164a111@grosbein.net> <a69d872a-f21e-de66-7677-58beccb0f023@otcnet.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
30.12.2020 16:44, Victor Gamov wrote: > Currently I'm thinking about ethernet flow control: Host-B connected to VLAN-750 on the third switch has 1G link (via igb driver) > and both Host-A and Host-B has fc=3. So when Host-B get microburst it can send PAUSE and Host-A start to fill queue. > But FC disabled for all switches and I'm not sure about PAUSE can be propagated from Host-A to Host-B AFAIK, pause frames are not forwarded. For short term you should make flow control settings consistent between link partners, maybe increase kern.ipc.maxsockbuf and then net.inet.udp.recvspace. But for long term you should consider adding more links and create LACP aggregate so not input nor output link be close to congestion.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6c780827-e764-8053-356b-a921e0892c15>