From johnm@advocap.org Tue Jul 5 12:31:46 2005 Received: with ECARTIS (v1.0.0; list fam); Tue, 05 Jul 2005 12:31:49 -0700 (PDT) Received: from oshkosh.advocap.org (oshkosh.advocap.org [216.170.138.63]) by oss.sgi.com (8.12.10/8.12.10/SuSE Linux 0.7) with ESMTP id j65JVhH9025095 for ; Tue, 5 Jul 2005 12:31:46 -0700 Received: from [192.168.1.19] (johnm.advocap.org [192.168.1.19]) by oshkosh.advocap.org (Postfix) with ESMTP id E54C32C069 for ; Tue, 5 Jul 2005 14:30:08 -0500 (CDT) Message-ID: <42CADFBE.7040105@advocap.org> Date: Tue, 05 Jul 2005 14:30:06 -0500 From: John McMonagle Organization: Advocap Inc User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.8) Gecko/20050513 Debian/1.7.8-1 X-Accept-Language: en MIME-Version: 1.0 To: fam@oss.sgi.com Subject: fam limit on number of monitor requests. Content-Type: multipart/mixed; boundary="------------000406050401050103030901" X-archive-position: 320 X-ecartis-version: Ecartis v1.0.0 Sender: fam-bounce@oss.sgi.com Errors-to: fam-bounce@oss.sgi.com X-original-sender: johnm@advocap.org Precedence: bulk X-list: fam This is a multi-part message in MIME format. --------------000406050401050103030901 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Am running fam in a kde ltsp environment using debian sarge. After about 10 people log in sessions start to stall. Was getting errors in syslog from famd saying it is out of file handles. Jun 27 13:28:52 oshkosh famd[5361]: failed to accept new client: Too many open files Jun 27 13:29:00 oshkosh famd[5361]: localclient socket(PF_UNIX, SOCK_STREAM, 0): Too many open files restarting famd makes things resume. I added to the startup of famd ulimit -n 8196 This stops the errors to syslog but still hangs when fam has a bit over 1024 files opened. I check with lsof. To cover the inevitable question ;-) /proc/sys/fs/file-max 2040670 /proc/sys/fs/file-nr 14220 0 2040670 From running the test program it seems to open another file for each directory or file, Tried rebuilding fam changing in DNotify.h enum { QUEUESIZE = 1024 }; to enum { QUEUESIZE = 8196 }; Didn't help :-( At the moment running without famd but would prefer to run it if possible. Personally I consider this to be a catastrophic bug :( My largest system could have 50 users logged in with kde so I'm guessing could get by with 4096 and 8196 would really safe. Next I'm planning on removing dnotify support and see what happens. Any thing else to try to fix this? John --------------000406050401050103030901 Content-Type: text/x-vcard; charset=utf-8; name="johnm.vcf" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="johnm.vcf" begin:vcard fn:John McMonagle n:McMonagle;John org:Advocap Inc adr;dom:;;2929 Harrison St;Oshkosh;WI;54936 email;internet:johnm@advocap.org title:IT Manager tel;work:920-426-0150 x-mozilla-html:FALSE url:http://www.advocap.org version:2.1 end:vcard --------------000406050401050103030901--