From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from relay.sgi.com (relay3.corp.sgi.com [198.149.34.15]) by oss.sgi.com (Postfix) with ESMTP id 4AB8B7CBF for ; Wed, 19 Jun 2013 18:02:20 -0500 (CDT) Received: from cuda.sgi.com (cuda1.sgi.com [192.48.157.11]) by relay3.corp.sgi.com (Postfix) with ESMTP id D2B60AC005 for ; Wed, 19 Jun 2013 16:02:16 -0700 (PDT) Received: from mx3-phx2.redhat.com (mx3-phx2.redhat.com [209.132.183.24]) by cuda.sgi.com with ESMTP id rGtKsSZBdwkBRFPl for ; Wed, 19 Jun 2013 16:02:15 -0700 (PDT) Date: Wed, 19 Jun 2013 19:02:13 -0400 (EDT) From: Nathan Scott Message-ID: <1040361622.4508678.1371682933909.JavaMail.root@redhat.com> In-Reply-To: References: <419435719.1662203.1371431489790.JavaMail.root@redhat.com> <20130617024603.GJ29338@dastard> <1597962722.1767244.1371447710942.JavaMail.root@redhat.com> <20130617111347.GL29338@dastard> Subject: Useful stats (was Re: definitions for /proc/fs/xfs/stat) MIME-Version: 1.0 Reply-To: Nathan Scott List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: xfs-bounces@oss.sgi.com Sender: xfs-bounces@oss.sgi.com To: Mark Seger Cc: xfs@oss.sgi.com Hi Mark, ----- Original Message ----- > ... > On a totally different topic, and if you like we can start a different > thread on it, I'd be interested in adding some monitoring stats to collectl > for xfs and could use some suggestions of what are the most important. Did you receive any other feedback on this (offlist perhaps)? I was kinda hoping someone else would chime in, but can share our experiences with PCP if it helps. We provide options to record all xfs.* metrics & an alternative to record just the core I/O stats (so log I/O and the file read/write metrics). With PCP one can record any subset at any rate, so those are just convenience / default suggestions. In terms of recording in collectl, not sure what the requirements/parameters are for you but I'd suggest recording all the data (in save-to-disk mode), and IIRC you've got brief/verbose reporting modes - I'd focus on the log, file I/O and buffer I/O stats when being "brief", and expand out to showing in-core inode/dquot stats, ail/transaction stats, dir/attr operation stats. You might consider another very-verbose kinda view with all of the various btree stats, but that may be going too far. Note that these are all aggregate statistics, there has been on-and-off discussion for many years about adding some per-filesystem XFS stats, but to date that has not happened - perhaps someday it will, so keep that in mind for collectl too I guess. Also, be aware there is a separate file for quota metrics, for hysterical raisins (/proc/fs/xfs/xqmstat). HTH. cheers. -- Nathan _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs