Hi Lucas. Thanks for the vote. As far as I understand, it should be enough to adjust `net.core.somaxconn` and `net.ipv4.tcp_max_syn_backlog` accordingly for allocating sufficient backlog size.
2021年10月18日(月) 17:37 Lucas Bradstreet <lu...@confluent.io.invalid>: > The other related kernel config that I can recall is > net.ipv4.tcp_max_syn_backlog. > > On Mon, Oct 18, 2021 at 4:32 PM Lucas Bradstreet <lu...@confluent.io> > wrote: > > > Thank you for the KIP. +1 (non-binding) > > > > For the implementation can we ensure that any kernel parameters that may > > need to also be adjusted are documented in the config documentation > > (e.g. net.core.somaxconn)? > > > > > > On Mon, Oct 18, 2021 at 4:23 PM Haruki Okada <ocadar...@gmail.com> > wrote: > > > >> Hi Luke. > >> > >> Thank you for the vote. > >> Updated KIP to link to ServerSocket#bind javadoc. > >> > >> > >> 2021年10月18日(月) 17:00 Luke Chen <show...@gmail.com>: > >> > >> > Hi Okada, > >> > Thanks for the KIP. > >> > +1 (non-binding) > >> > > >> > One thing to add is that you should add ServerSocket#bind java doc > link > >> > into the KIP. > >> > I don't think everyone is familiar with the definition of the method > >> > parameters. > >> > > >> > Thank you. > >> > Luke > >> > > >> > On Mon, Oct 18, 2021 at 3:43 PM Haruki Okada <ocadar...@gmail.com> > >> wrote: > >> > > >> > > Hi Kafka. > >> > > > >> > > Let me bump this VOTE thread for the KIP. > >> > > We applied proposed changes in the KIP to our large Kafka cluster by > >> > > building patched Kafka internally and confirmed it's working well. > >> > > > >> > > Please feel free to give your feedback if there's any points to be > >> > > clarified in the KIP. > >> > > > >> > > Thanks, > >> > > > >> > > 2021年8月9日(月) 11:25 Haruki Okada <ocadar...@gmail.com>: > >> > > > >> > > > Thanks for your comment LI-san. > >> > > > > >> > > > Could anyone else review and vote for the KIP? > >> > > > > >> > > > I think the situation described in the KIP's motivation can happen > >> in > >> > any > >> > > > large-scale Kafka deployment, so may be helpful for many users > while > >> > the > >> > > > proposed changes are small enough. > >> > > > > >> > > > > >> > > > Thanks, > >> > > > > >> > > > 2021年8月3日(火) 15:49 Xiangyuan LI <flashmouse2...@gmail.com>: > >> > > > > >> > > >> Hi Haruki Okada: > >> > > >> i read your comment, thx for your detail explain! > >> > > >> add backlog parameter is a useful suggestion, hope it could > >> added to > >> > > >> kafka. > >> > > >> > >> > > >> Haruki Okada <ocadar...@gmail.com> 于2021年8月2日周一 上午7:43写道： > >> > > >> > >> > > >> > Hi, Kafka. > >> > > >> > > >> > > >> > I would like to start a vote on KIP that makes SocketServer > >> > acceptor's > >> > > >> > backlog size configurable. > >> > > >> > > >> > > >> > KIP: > >> > > >> > > >> > > >> > > >> > > >> > >> > > > >> > > >> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-764%3A+Configurable+backlog+size+for+creating+Acceptor > >> > > >> > > >> > > >> > Discussion thread: > >> > > >> > > >> > > >> > > >> > > >> > >> > > > >> > > >> > https://lists.apache.org/thread.html/rd77469b7de0190d601dd37bd6894e1352a674d08038bcfe7ff68a1e0%40%3Cdev.kafka.apache.org%3E > >> > > >> > > >> > > >> > Thanks, > >> > > >> > > >> > > >> > -- > >> > > >> > ======================== > >> > > >> > Okada Haruki > >> > > >> > ocadar...@gmail.com > >> > > >> > ======================== > >> > > >> > > >> > > >> > >> > > > > >> > > > > >> > > > -- > >> > > > ======================== > >> > > > Okada Haruki > >> > > > ocadar...@gmail.com > >> > > > ======================== > >> > > > > >> > > > >> > > > >> > > -- > >> > > ======================== > >> > > Okada Haruki > >> > > ocadar...@gmail.com > >> > > ======================== > >> > > > >> > > >> > >> > >> -- > >> ======================== > >> Okada Haruki > >> ocadar...@gmail.com > >> ======================== > >> > > > -- ======================== Okada Haruki ocadar...@gmail.com ========================
