Date: Fri, 26 Jul 2019 15:19:42 +0200 From: Robert Schulze <rs@bytecamp.net> To: freebsd-questions@freebsd.org Subject: Re: Help:: Listen queue overflow killing servers Message-ID: <a9d0ee57-ee94-8cff-a711-2c03fc55ffb7@bytecamp.net> In-Reply-To: <a4675162-4241-2145-3380-c12253032da1@ifdnrg.com> References: <3a62375a-432c-3533-a7bc-e5573c26fa9c@ifdnrg.com> <92866b76-5f11-2523-cc8f-0d92cc91a50e@bytecamp.net> <a4675162-4241-2145-3380-c12253032da1@ifdnrg.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, Am 26.07.19 um 15:08 schrieb Paul Macdonald via freebsd-questions: > 1) why are the queues filling up (i'll revert to seperate logs to see > if that helps, although the issue is sporadic, and first time on this box) like I already mentioned: if a webserver process is stuck in disk-wait, this process will probably not answer any further requests, so the listen backlog will fill up. > 2) Once the queues are over limit, is this actually unresolvable other > than a hard reboot? again: your problem are the webserver processes in disk-wait. This process state is uninterruptible, processes in this state cannot be killed. A process simply reaching somaxcon will not interfere with reboot/kill. > I've increased the queue size via > > sysctl kern.ipc.somaxconn > > to 1024 for now You will not solve your problem by increasing somaxconn. You have to figure out why these processes are stuck in disk i/o (see previous hints about procstat). regards Robert Schulze
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?a9d0ee57-ee94-8cff-a711-2c03fc55ffb7>