Return-Path: linux-nfs-owner@vger.kernel.org Received: from mx1.redhat.com ([209.132.183.28]:60952 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934805Ab3FSOHN (ORCPT ); Wed, 19 Jun 2013 10:07:13 -0400 From: David Howells In-Reply-To: <1371649292.5522.1.camel@leira.trondhjem.org> References: <1371649292.5522.1.camel@leira.trondhjem.org> <20130618201248.6cc88501@csupomona.edu> To: "Myklebust, Trond" Cc: dhowells@redhat.com, Brian De Wolf , Linux NFS list Subject: Re: Issues using new idmapper in large sites Date: Wed, 19 Jun 2013 15:07:10 +0100 Message-ID: <7562.1371650830@warthog.procyon.org.uk> Sender: linux-nfs-owner@vger.kernel.org List-ID: Myklebust, Trond wrote: > > It's fast but...it also missed most of my users (it only has 503 > > cached, there are 36287 total). The refcount and number of keys > > skyrocket even further on repeated runs but the number of missing > > users remains the same. Okay... I need to improve keyring capacity. Current max keyring capacity is (PAGE_SIZE - header) / sizeof(key *) in other words, 500-ish on a 64-bit machine. Let me have a play - in theory, it shouldn't be too hard; the main trick is retaining RCU-based reading and, ideally, tree size. Plus if I can sort it too... David