From owner-freebsd-fs@FreeBSD.ORG Thu Apr 2 19:25:58 2015 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id AE97CE19 for ; Thu, 2 Apr 2015 19:25:58 +0000 (UTC) Received: from mail.tezzaron.com (mail.tezzaron.com [50.206.41.178]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 727EE7A4 for ; Thu, 2 Apr 2015 19:25:58 +0000 (UTC) Received: from delaware.tezzaron.com ([10.252.50.1]) by mail.tezzaron.com (IceWarp 11.1.2.0 x64) with ASMTP (SSL) id 201504021425572207; Thu, 02 Apr 2015 14:25:57 -0500 Message-ID: <551D97C5.8020405@tezzaron.com> Date: Thu, 02 Apr 2015 14:25:57 -0500 From: Adam Guimont User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.6.0 MIME-Version: 1.0 To: rick@havokmon.com Subject: Re: NFSD high CPU usage References: <20150402105040.Horde.DpcVnMHXCV_MvaXmGcnU1g8@www.vfemail.net> In-Reply-To: <20150402105040.Horde.DpcVnMHXCV_MvaXmGcnU1g8@www.vfemail.net> Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Cc: freebsd-fs@freebsd.org X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 02 Apr 2015 19:25:58 -0000 Rick Romero wrote: > Are the disks busy before it happens? I'm far from an expert, but when > running ZFS with NFS, I've had a lot of issues. My final resolutions were > to turn ASYNC off and have log devices and I even have SSD volumes now. > Otherwise under load the NFS server gets hung up. It never seemed to happen > on UFS, but due to the number of small files I have, ZFS provides the best > backup functionality. I'm now trying to move all functions from NFS (to > more TCP client/server). > > You have different info than I've gathered, and it might be because of > usage. I actively use the system that I've seen NFS dump on, so I see the > slowness beginning. Once NFS dies, the drive load goes back to normal. I > wonder, if maybe you are just managing a system for others, and you don't > see it until after the fact? Just a thought based on my limited > experience. No, the disks are not busy before this happens. I use the server every day and keep a pretty close eye on it. The disks can get busy but it doesn't spike nfsd that much and usually doesn't last more than a few seconds. When this particular issue happens with the nfsd CPU spike, it will last until the job running on the client workstation gets killed or when the client workstation is rebooted. After that it takes a few seconds for the TCP buffers to flush out and allow other clients to connect again. Regards, Adam Guimont