Re: procmail keeps dieing on freebsd 5.4 with postfix
Matt Juszczak wrote: Hi all, We had a mail server running with FreeBSD 5.4, about 3,000 accounts, and postfix. Recently, I turned procmail on in postfix (mailbox_command=/usr/procmail) and the machine has been locking up weekly ever since. And when this machines crashes, it crashes hard ... and procmail is always on the screen as the error causer when it happens. Patient: Doctor, my arm hurts when I do this. Doctor: Then stop doing that. Seriously though, you need to provide some more detailed information if you want anyone here to be able to help you. Start with explaining why you decided to change MDAs in the first place since I'm sure I'm not the only one thinking you must be nuts to make such a major change on a production system with a potential 3000-user lynch mob waiting in the wings. What were you using for local delivery before this? Was there a problem with it or were you looking for new features, etc.? I know you all want messages, but I never seem to be here and my co workers reboot the box on me to fix it. Both times; however, we've had to run fsck from single user mode and also refresh the postfix queue. If you're not around to see the console messages how do you know procmail is always the error causer? Perhaps this is conveyed to you by your co-workers but if so, why don't they tell you the complete error message so you can convey it to us? Leaving that aside, however, what about the logs? Certainly /var/log/maillog should provide some clues if the problem is really your MDA (more on this below). Also we'd need to know something about your configuration (i.e. contents of main.cf and master.cf for starters) to help you with a MTA/MDA problem. Does anyone have any ideas why procmail could be causing my system to completely hard lock every other couple of days? I disabled procmail for now and I know (knock on wood) the machine should be fine like it used to be ... but all these hard locks could eventually drive the freebsd box mad, and I wouldn't want to do a reinstall. FWIW this doesn't sound like a software issue (except maybe a massive memory leak(??)) but then again, I'm saying this with very little useful information provided by you. Have you done any basic hardware checks (e.g. memtest, case and cpu cooling, power supply integrity, etc.)? You've stated that these lock-ups occur every week at the beginning of your post then you say later it's every couple of days. Which is it? Also, please try to precisely define locking up and crashes. It's unclear to me based on your description and the (possibly misleading) subject line what portions of the system are affected. Precision matters IMHO. Cheers, G ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: procmail keeps dieing on freebsd 5.4 with postfix
Appreciate the response :) Here's my message the way it should have been originally Seriously though, you need to provide some more detailed information if you want anyone here to be able to help you. Start with explaining why you decided to change MDAs in the first place since I'm sure I'm not the only one thinking you must be nuts to make such a major change on a production system with a potential 3000-user lynch mob waiting in the wings. What were you using for local delivery before this? Was there a problem with it or were you looking for new features, etc.? We are currently moving to a new mail server that is FreeBSD-based. Our old mail server is a chrooted slackware box that hasn't been upgraded in years because no one even had access to it for a while (the management of the company I work for used to stink, its better now). Our new mail server has 3000 accounts on it, that are active, but only about 50 of them are actually functioning (one of our virtual domains). We haven't switched the MX record for our main ISP yet, we're waiting to make sure the box is stable first. So to answer your question, there is only about a 50-user lynch mob and most of those users are internal to our ISP (employees, etc.) I would not make a change on something that had more live users, especially paying customers. Our current mail server supports procmail, and we have about 50 users who use it. Therefore, thats why I was turning it on on the new server. We're working on basically mirroring the old server to the new one and making sure that our change will be swift and efficient. I've considered using postfix's internal LDA and just calling procmail from inside a .forward file for those users who need it/want it ... this might end up fixing the problems. If you're not around to see the console messages how do you know procmail is always the error causer? Perhaps this is conveyed to you by your co-workers but if so, why don't they tell you the complete error message so you can convey it to us? Leaving that aside, however, what about the logs? Certainly /var/log/maillog should provide some clues if the problem is really your MDA (more on this below). Also we'd need to know something about your configuration (i.e. contents of main.cf and master.cf for starters) to help you with a MTA/MDA problem. Its happened twice now. The first time this problem happened was late at night, about 2 days after I made the change to the LDA. The machine would not respond to ping, and nagios was alerting us like crazy that the box was down. The machine was non-responsive to the keyboard, and the console had a dump on it, about 15 lines long, with procmail written all over it. I turned procmail off after rebooting the machine, running fsck, restoring postfix to a functioning state, etc. Procmail remained disabled for about three weeks, in which the box ran fine. Yesterday afternoon we switched the LDA back to procmail, and the machine ran fine over night. On my way into work today, I got paged that the box was down from nagios and called. The tech that was here rebooted the machine, but before he did he said, in his own words There was a bunch of crap on the screen with procmail this and procmail that, and the machine was locked hard.. I've disabled procmail again and it seems to be running stable. As far as logs, nothing the maillog cuts out at 11:14 AM and cuts back in at 11:21 AM, with no errors in between. FWIW this doesn't sound like a software issue (except maybe a massive memory leak(??)) but then again, I'm saying this with very little useful information provided by you. Have you done any basic hardware checks (e.g. memtest, case and cpu cooling, power supply integrity, etc.)? Yes, the machine has been checked. We ran memtest on it, etc., with no problems. The machine is about 2 months old; however, so its passed its burn in test but could have issues, but I doubt thats the problem. You've stated that these lock-ups occur every week at the beginning of your post then you say later it's every couple of days. Which is it? Also, please try to precisely define locking up and crashes. It's unclear to me based on your description and the (possibly misleading) subject line what portions of the system are affected. Precision matters IMHO. See above. Its occured twice in a one month span but most of that time procmail was not running. It occurs usually within 24-48 hours of switching procmail back on. Thanks, hope this helps a little more! -Matt ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: procmail keeps dieing on freebsd 5.4 with postfix
OK, here's the funny thing. We did a mail flood test, and our mail server stood up fine, but our LDAP server (which was handling all the queries) ended up crashing with a similar message ... so now I've got two machines running 5.4 with the same behavior. Here's the message. Remember, this is on the LDAP machine, not the Postfix/Procmail machine, but the error we received earlier was similar. Kernel Trap 12 with interrupts disabled Fatal trap 12: page fault while in kernel mode CPUID=1, apic ID=00 fault virtual address = 0x24 fault code = supervisor read, page not present instruction pointer = 0x8:0xc6644eff stack pointer = 0x10: 0xdaa86b48 frame pointer = 0x10 :0xdae86b5c code segment: base 0x0 limit 0xf type 0x1d, can't read my handwriting here def32, 1 processes eflags = resume, IOPL=0 current process = 44091 (slapd) trap number = 12 panic page fault cpuid = 1 Hope that helps. I dont think FreeBSD should crash like this because it was getting hit hard with queries . but I could be wrong. -Matt ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: procmail keeps dieing on freebsd 5.4 with postfix
On Wed, 15 Jun 2005, Matt Juszczak wrote: We are currently moving to a new mail server that is FreeBSD-based. Our old mail server is a chrooted slackware box that hasn't been upgraded in years because no one even had access to it for a while (the management of the company I work for used to stink, its better now). Our new mail server has 3000 accounts on it, that are active, but only about 50 of them are actually functioning (one of our virtual domains). We haven't switched the MX record for our main ISP yet, we're waiting to make sure the box is stable first. So to answer your question, there is only about a 50-user lynch mob and most of those users are internal to our ISP (employees, etc.) I would not make a change on something that had more live users, especially paying customers. Our current mail server supports procmail, and we have about 50 users who use it. Therefore, thats why I was turning it on on the new server. We're working on basically mirroring the old server to the new one and making sure that our change will be swift and efficient. I've considered using postfix's internal LDA and just calling procmail from inside a .forward file for those users who need it/want it ... this might end up fixing the problems. Did you build procmail from ports or bare? Just asking because there are ten patch files included with the port... -Warren Block * Rapid City, South Dakota USA ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: procmail keeps dieing on freebsd 5.4 with postfix
Ports ... I always use ports. On Wed, 15 Jun 2005, Warren Block wrote: On Wed, 15 Jun 2005, Matt Juszczak wrote: We are currently moving to a new mail server that is FreeBSD-based. Our old mail server is a chrooted slackware box that hasn't been upgraded in years because no one even had access to it for a while (the management of the company I work for used to stink, its better now). Our new mail server has 3000 accounts on it, that are active, but only about 50 of them are actually functioning (one of our virtual domains). We haven't switched the MX record for our main ISP yet, we're waiting to make sure the box is stable first. So to answer your question, there is only about a 50-user lynch mob and most of those users are internal to our ISP (employees, etc.) I would not make a change on something that had more live users, especially paying customers. Our current mail server supports procmail, and we have about 50 users who use it. Therefore, thats why I was turning it on on the new server. We're working on basically mirroring the old server to the new one and making sure that our change will be swift and efficient. I've considered using postfix's internal LDA and just calling procmail from inside a .forward file for those users who need it/want it ... this might end up fixing the problems. Did you build procmail from ports or bare? Just asking because there are ten patch files included with the port... -Warren Block * Rapid City, South Dakota USA !DSPAM:42b076c8956801608011501! ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
