Received: (qmail 12719 invoked by uid 2012); 19 Nov 1997 11:15:48 -0000 Message-Id: <19971119111548.12718.qmail@hyperreal.org> Date: 19 Nov 1997 11:15:48 -0000 From: Ray Bellis Reply-To: rpb@community.net.uk To: apbugs@hyperreal.org Subject: CGIs run using 'rsh' frequently abort prematurely. X-Send-Pr-Version: 3.2 >Number: 1434 >Category: os-osf >Synopsis: CGIs run using 'rsh' frequently abort prematurely. >Confidential: no >Severity: serious >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Wed Nov 19 03:20:01 PST 1997 >Last-Modified: Tue Aug 4 19:15:27 PDT 1998 >Originator: rpb@community.net.uk >Organization: >Release: 1.2.4 >Environment: OSF1 alpha01.community.net.uk V3.2 41 alpha DEC cc. >Description: So as not to run 'analog' on our main server our CGIs indirectly call 'rsh' to offload the job onto another server. Since upgrading to 1.2.4 we've had problems where reports would abort prematurely, with no HTML document returned. Log symptoms would include lines such as 'rsh: select: bad file number', indicating that the Apache server had closed the output socket. >How-To-Repeat: >Fix: Defining USE_FLOCK_SERIALIZED_ACCEPT (although supposedly only related to problems on new listening sockets) appears to have cured the problem >Audit-Trail: State-Changed-From-To: open-analyzed State-Changed-By: marc State-Changed-When: Wed Nov 19 20:17:22 PST 1997 State-Changed-Why: I do find it somewhat odd that USE_FLOCK_SERIALIZED_ACCEPT would do anything. Are you sure nothing else changed when you did this? Is it easy to reproduce the problem? ie. even if it doesn't happen every other time, can you always make it happen by making a number of requests? Do you have a lot of virtual hosts? How many total Listen directives + logfiles do you have? Could you be running out of file descriptors? It is possible rsh is picky and doesn't like high file descriptors. From: Marc Slemko To: Apache bugs database Cc: Subject: Re: os-osf/1434: CGIs run using 'rsh' frequently abort prematurely. (fwd) Date: Thu, 20 Nov 1997 07:27:43 -0700 (MST) ---------- Forwarded message ---------- Date: Thu, 20 Nov 1997 09:02:32 +0000 (GMT Standard Time) From: Ray Bellis To: marc@hyperreal.org Cc: apache-bugdb@apache.org, marc@apache.org Subject: Re: os-osf/1434: CGIs run using 'rsh' frequently abort prematurely. On 20 Nov 1997 marc@hyperreal.org wrote: > Synopsis: CGIs run using 'rsh' frequently abort prematurely. > > State-Changed-From-To: open-analyzed > State-Changed-By: marc > State-Changed-When: Wed Nov 19 20:17:22 PST 1997 > State-Changed-Why: > I do find it somewhat odd that USE_FLOCK_SERIALIZED_ACCEPT > would do anything. > > Are you sure nothing else changed when you did this? There shouldn't have been any other changes, it was exactly the same build tree as I normally use but with the #define added in conf.h. > Is it easy to reproduce the problem? ie. even if it doesn't > happen every other time, can you always make it happen > by making a number of requests? In order to determine the answer to some of your questions I've just started the old binary back up and am now unable to duplicate the problem. I'll monitor the situtation and see if we can determine at what point the problem reoccurs. When the problem was apparent we found that one request would work, and then we wouldn't be able to make the script work for ages, and then it would just start working again for a few requests. > Do you have a lot of virtual hosts? How many total > Listen directives + logfiles do you have? Could you > be running out of file descriptors? It is possible > rsh is picky and doesn't like high file descriptors. I'm currently running 159 virtual servers, 'lsof' reports about 172 fds per process. I had considered that lack of file descriptors might be a problem but Digital Unix has a getdtablesize() of 4096 and I also tested that sizeof(fd_set) is actually big enough to hold that many fds (it is). I've only got two listen directives, a global 'Listen 80' and a single-IP 'Listen xxx.xxx.xxx.xxx:443' for a single SSL virtual server. Ray. -- Ray Bellis, MA(Oxon) - Technical Manager - Oxford CommUnity Internet plc Windsor House, 12 High Street, Kidlington, OXFORD OX5 2PJ UK Telephone: +44-1865-856000 Fax: +44-1865-856001 Email: ray.bellis@community.net.uk URL: http://www.community.co.uk/ Release-Changed-From-To: 1.2.4 (+algroup SSL patches)-1.2.4 Release-Changed-By: coar Release-Changed-When: Mon Nov 24 07:29:01 PST 1997 State-Changed-From-To: analyzed-feedback State-Changed-By: coar State-Changed-When: Thu May 28 12:14:39 PDT 1998 State-Changed-Why: [This is a standard response.] This Apache problem report has not been updated recently. Please reply to this message if you have any additional information about this issue, or if you have answers to any questions that have been posed to you. If there are no outstanding questions, please consider this a request to try to reproduce the problem with the latest software release, if one has been made since last contact. If we don't hear from you, this report will be closed. State-Changed-From-To: feedback-closed State-Changed-By: coar State-Changed-When: Tue Aug 4 19:15:27 PDT 1998 State-Changed-Why: [This is a standard response.] No response from submitter, assuming issue has been resolved. >Unformatted: [In order for any reply to be added to the PR database, ] [you need to include in the Cc line ] [and leave the subject line UNCHANGED. This is not done] [automatically because of the potential for mail loops. ]