Received: (qmail 5476 invoked by uid 2012); 26 Mar 1998 17:04:38 -0000 Message-Id: <19980326170438.5475.qmail@hyperreal.org> Date: 26 Mar 1998 17:04:38 -0000 From: Jason Priebe Reply-To: priebe@wral-tv.com To: apbugs@hyperreal.org Subject: unkillable httpd processes X-Send-Pr-Version: 3.2 >Number: 2002 >Category: os-irix >Synopsis: unkillable httpd processes >Confidential: no >Severity: critical >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Thu Mar 26 09:10:00 PST 1998 >Last-Modified: Sat Nov 14 08:34:20 PST 1998 >Originator: priebe@wral-tv.com >Organization: >Release: 1.2.5 >Environment: IRIX barkley 6.2 03131015 IP22 gcc version 2.7.2.3 (running on another machine under IRIX 5.3) >Description: To restart our httpd, we use kill -HUP `cat /usr/local/etc/httpd/logs/httpd.pid` Most of the time, this successfully restarts the web server, allowing us to rotate our logs. But sometimes the server shuts down leaving a few httpd processes that are unkillable. (they refuse to respond to "kill -9"). The only recourse at that point is a reboot. We have also seen similar behavior when stopping the web server, using: kill `cat /usr/local/etc/httpd/logs/httpd.pid` Some of the processes hang inexplicably. The server root is on an NFS-mounted volume, if that is of any help. >How-To-Repeat: >Fix: n >Audit-Trail: From: Marc Slemko To: Jason Priebe Cc: apbugs@hyperreal.org Subject: Re: os-irix/2002: unkillable httpd processes Date: Thu, 26 Mar 1998 10:52:07 -0700 (MST) Did you try adding a "LockFile /tmp/accept.lock" or something along those lines to your Configuration file? You must not have it on a NFS mounted disk. On 26 Mar 1998, Jason Priebe wrote: > > >Number: 2002 > >Category: os-irix > >Synopsis: unkillable httpd processes > >Confidential: no > >Severity: critical > >Priority: medium > >Responsible: apache > >State: open > >Class: sw-bug > >Submitter-Id: apache > >Arrival-Date: Thu Mar 26 09:10:00 PST 1998 > >Last-Modified: > >Originator: priebe@wral-tv.com > >Organization: > apache > >Release: 1.2.5 > >Environment: > IRIX barkley 6.2 03131015 IP22 > gcc version 2.7.2.3 (running on another machine under IRIX 5.3) > >Description: > To restart our httpd, we use > > kill -HUP `cat /usr/local/etc/httpd/logs/httpd.pid` > > Most of the time, this successfully restarts the web server, allowing us > to rotate our logs. But sometimes the server shuts down leaving a few > httpd processes that are unkillable. (they refuse to respond to "kill -9"). > The only recourse at that point is a reboot. > > We have also seen similar behavior when stopping the web server, using: > > kill `cat /usr/local/etc/httpd/logs/httpd.pid` > > Some of the processes hang inexplicably. The server root is on an > NFS-mounted volume, if that is of any help. > >How-To-Repeat: > > >Fix: > n > >Audit-Trail: > >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. ] > > > State-Changed-From-To: open-closed State-Changed-By: brian State-Changed-When: Wed May 20 04:22:24 PDT 1998 State-Changed-Why: [This is a standard response.] No response from submitter, assuming issue has been resolved. From: "Jason S. Priebe" To: brian@hyperreal.org, apbugs@apache.org Cc: Subject: Re: os-irix/2002: unkillable httpd processes Date: Wed, 20 May 1998 08:09:00 -0400 (EDT) -----BEGIN PGP SIGNED MESSAGE----- On 20 May 1998 brian@hyperreal.org wrote: > Synopsis: unkillable httpd processes > > State-Changed-From-To: open-closed > State-Changed-By: brian > State-Changed-When: Wed May 20 04:22:24 PDT 1998 > State-Changed-Why: > [This is a standard response.] > No response from submitter, assuming issue has been resolved. I didn't realize I was supposed to respond to anything. The issue has not been resolved. We've had to hack together a solution where we try to start the server every minute for 15 minutes after we restart the server just in case it doesn't restart properly. - -------------------------------------------------- Jason Priebe WRAL OnLine priebe@wral-tv.com Raleigh, NC see http://www.wral-tv.com/~priebe/ for public key -----BEGIN PGP SIGNATURE----- Version: 2.6.2 iQCVAwUBNWLH4eqkj/9w/z2hAQEzwQP/bGDmcaVwD9RZHf6w5OdhkyxrluD8byoI PJEZoyC5Bad5sWGuiJ/7Jk2qWcvxRVD79kOPethdLyQN1gUnT4KaB70s5CXw/Btt H84ItC2qcgVKe7i1O8P01CEp8OQTHD1q457L692wkUA4Sd5BKzen2PGm/hOz4GVo RElYr87Jgo0= =FyTq -----END PGP SIGNATURE----- From: Brian Behlendorf To: "Jason S. Priebe" Cc: apbugs@apache.org Subject: Re: os-irix/2002: unkillable httpd processes Date: Wed, 20 May 1998 06:46:15 -0700 (PDT) On Wed, 20 May 1998, Jason S. Priebe wrote: > I didn't realize I was supposed to respond to anything. The issue > has not been resolved. We've had to hack together a solution where > we try to start the server every minute for 15 minutes after we > restart the server just in case it doesn't restart properly. Back in March Marc asked you: > Did you try adding a "LockFile /tmp/accept.lock" or something along > those lines to your Configuration file? You must not have it on a NFS > mounted disk. There was no response recorded in the bug database, and since you said your serverroot was on an NFS volume this was almost certainly the problem. Did you get his message? Was that the problem? Brian From: "Jason S. Priebe" (by way of Brian Behlendorf ) To: apbugs@apache.org Cc: Subject: Re: os-irix/2002: unkillable httpd processes Date: Wed, 20 May 1998 15:14:09 -0700 -----BEGIN PGP SIGNED MESSAGE----- On Wed, 20 May 1998, Brian Behlendorf wrote: > > Did you try adding a "LockFile /tmp/accept.lock" or something along > > those lines to your Configuration file? You must not have it on a NFS > > mounted disk. > > There was no response recorded in the bug database, and since you said > your serverroot was on an NFS volume this was almost certainly the > problem. Did you get his message? Was that the problem? Our documents are on an NFS volume, but the ServerRoot is on a local disk. - -------------------------------------------------- Jason Priebe WRAL OnLine priebe@wral-tv.com Raleigh, NC see http://www.wral-tv.com/~priebe/ for public key -----BEGIN PGP SIGNATURE----- Version: 2.6.2 iQCVAwUBNWLfY+qkj/9w/z2hAQF/dAQAvZN6eYa9cPGQ5aQ06wqcSr3987ELX9B2 TLbx41+ylLtMbrQ8peFD3icbSyDeHY76CtAcfnFUY2iKfZiDcXzb/U/es7wRyVif MOxuwVIpPUBsuJDVKBHry8oqkN+fSODR6BG5omNPi23v8dw1ew/o2YHAlIjBQKfK m2+jJRqCzS0= =XDu4 -----END PGP SIGNATURE----- State-Changed-From-To: closed-feedback State-Changed-By: brian State-Changed-When: Wed May 20 15:12:01 PDT 1998 State-Changed-Why: user reports still having the problem, re-open the report. From: Brian Behlendorf To: priebe@wral-tv.com, apbugs@apache.org Cc: Subject: Re: os-irix/2002: unkillable httpd processes Date: Wed, 20 May 1998 15:20:03 -0700 At 09:49 AM 5/20/98 -0400, you wrote: > Our documents are on an NFS volume, but the ServerRoot is on a local > disk. Hmm, okay, that's usually been the cause of unkillable processes that we've seen. To be fair, there /should/ be nothing that Apache can do to cause unkillable processes; that's clearly an operating system bug. The best we can hope to do is avoid some legitimate sequence of events which leads to that situation; so to help you any further we'd need some better information as to what conditions cause these sporadic unkillable processes occur. Not to sound like I'm avoiding the issue, but isn't Irix up to 6.4 now, or at least 6.3? If you can't kill a process you probably can't ktrace it. Can you see what state it was last in? I'm not familiar with Irix's process debugging and management tools. Brian --=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-- pure chewing satisfaction brian@apache.org brian@hyperreal.org From: Marc Slemko To: "Jason S. Priebe" Cc: Apache bugs database Subject: Re: os-irix/2002: unkillable httpd processes Date: Wed, 20 May 1998 19:15:26 -0600 (MDT) On 20 May 1998, Jason S. Priebe wrote: > The following reply was made to PR os-irix/2002; it has been noted by GNATS. > > From: "Jason S. Priebe" (by way of Brian Behlendorf ) > To: apbugs@apache.org > Cc: Subject: Re: os-irix/2002: unkillable httpd processes > Date: Wed, 20 May 1998 15:14:09 -0700 > > -----BEGIN PGP SIGNED MESSAGE----- > > On Wed, 20 May 1998, Brian Behlendorf wrote: > > > > Did you try adding a "LockFile /tmp/accept.lock" or something along > > > those lines to your Configuration file? You must not have it on a NFS > > > mounted disk. > > > > There was no response recorded in the bug database, and since you said > > your serverroot was on an NFS volume this was almost certainly the > > problem. Did you get his message? Was that the problem? > > Our documents are on an NFS volume, but the ServerRoot is on a local > disk. I am really tempted to say that this is an IRIX problem related to NFS. I regularily see unkillable processes (blocked in uninterruptable NFS things) on several versions of IRIX when NFS is involved; not just Apache, but tons of things. I really doubt this is Apache's fault; the only thing I can suggest is that you try upgrading and downgrading IRIX and adding and removing patches. From: Marc Slemko To: Apache bugs database Cc: Subject: Re: os-irix/2002: unkillable httpd processes (fwd) Date: Thu, 21 May 1998 10:01:51 -0600 (MDT) ---------- Forwarded message ---------- Date: Thu, 21 May 1998 08:28:58 -0400 (EDT) From: "Jason S. Priebe" To: Marc Slemko Subject: Re: os-irix/2002: unkillable httpd processes -----BEGIN PGP SIGNED MESSAGE----- On Wed, 20 May 1998, Marc Slemko wrote: > I am really tempted to say that this is an IRIX problem related to NFS. It could be -- since we run little else on the machine other than our httpd, we haven't seen such behavior manifested elsewhere. > I regularily see unkillable processes (blocked in uninterruptable NFS > things) on several versions of IRIX when NFS is involved; not just Apache, > but tons of things. I really doubt this is Apache's fault; the only thing > I can suggest is that you try upgrading and downgrading IRIX and adding > and removing patches. Thanks for the advice. I inherited these SGI boxes from my predecessor. If I had my druthers, we'd be running all Linux (even if it would take two or three times the number of machines to match the horsepower of the Challenges). - -------------------------------------------------- Jason Priebe WRAL OnLine priebe@wral-tv.com Raleigh, NC see http://www.wral-tv.com/~priebe/ for public key -----BEGIN PGP SIGNATURE----- Version: 2.6.2 iQCVAwUBNWQeDeqkj/9w/z2hAQHAbAP/UF/S1A8eHJMJl++YKq4xhWF+veTHm98u byBWL6eKA7g0I537ZkaTcgNMVeTJLzpyfra7pRGiBnYL77tAwb3p7pdxZLtAeCo2 xMTyQSfKqvnWUbJcu8/OWw4zXxxgntK7XNznZu+5OYig35OascvKc/gYLKkHhh4Y wsk8pZ6Lvig= =0FOW -----END PGP SIGNATURE----- From: Brian Behlendorf To: "Jason S. Priebe" Cc: apbugs@apache.org Subject: Re: os-irix/2002: unkillable httpd processes (fwd) Date: Thu, 21 May 1998 11:01:38 -0700 At 05:00 PM 5/21/98 -0000, Marc Slemko wrote: > > I am really tempted to say that this is an IRIX problem related to NFS. > > It could be -- since we run little else on the machine other than > our httpd, we haven't seen such behavior manifested elsewhere. An easy way to test it, depending on your available resources, would be to have a local non-NFS'd copy of your document tree. If you do that and don't have errors, well there's your answer. Brian --=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-- pure chewing satisfaction brian@apache.org brian@hyperreal.org Comment-Added-By: coar Comment-Added-When: Mon Sep 7 06:25:03 PDT 1998 Comment-Added: [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: lars State-Changed-When: Sat Nov 14 08:34:20 PST 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. ]