Mailing List - people say their e-mails "never appear"
Hi all,
I just installed Mailman on our load balanced Postfix servers. Both servers share a common NFS point.
Please forgive me, as I'm very new to Mailman. While I know my way around Postfix, troubleshooting Mailman is another issue. I can see that e-mail gets sent and delivered to the particular mailing list in question, but sometimes people tell me their e-mails never get posted to the mailing list. Other people say they submit their post, and it eventually gets there, after 10 or so minutes.
Since I don't know how to troubleshoot Mailman, I can only suspect that is has something to do with the NFS, although another part of me is telling me that it can't be.
Can you folks push me in the right direction? What should I be troubleshooting? Is there anything I can provide so that you folks can help me troubleshoot further?
Thanks in advance.
Patrick
On 9/5/07, Patrick M. wrote:
I just installed Mailman on our load balanced Postfix servers. Both servers share a common NFS point.
Did you install it on all of them, or just one? When you say they share a common NFS point, are you trying to use NFS for the postfix mail queue (which is a really, really bad idea), or just for the user mailboxes (which is also usually a bad idea, although it can be made somewhat less bad if you choose a mailbox format that is less NFS-unfriendly)?
Mailman itself is about as NFS-friendly as can be reasonably done, but that still leaves a lot of room for problems. Many of the other parts of most mail systems are much, much less NFS-friendly than Mailman.
Please forgive me, as I'm very new to Mailman. While I know my way around Postfix, troubleshooting Mailman is another issue.
When troubleshooting Mailman, you generally want to start with the steps outlined in FAQ 4.78 at <http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq04.078.htp>. You should also search the FAQ Wizard for other things that might be related to your problems, and you should also search the archives.
Since I don't know how to troubleshoot Mailman, I can only suspect that is has something to do with the NFS, although another part of me is telling me that it can't be.
If NFS is involved anywhere in the picture, odds of Mr. Murphy raising his nasty head have been increased by many, many orders of magnitude.
That said, there are occasionally times when such a system has problems that are actually unrelated to NFS.
-- Brad Knowles <brad@shub-internet.org> LinkedIn Profile: <http://tinyurl.com/y8kpxu>
Brad Knowles wrote:
On 9/5/07, Patrick M. wrote:
I just installed Mailman on our load balanced Postfix servers. Both servers share a common NFS point.
Did you install it on all of them, or just one? When you say they share a common NFS point, are you trying to use NFS for the postfix mail queue (which is a really, really bad idea), or just for the user mailboxes (which is also usually a bad idea, although it can be made somewhat less bad if you choose a mailbox format that is less NFS-unfriendly)?
Mailman itself is about as NFS-friendly as can be reasonably done, but that still leaves a lot of room for problems. Many of the other parts of most mail systems are much, much less NFS-friendly than Mailman.
Please forgive me, as I'm very new to Mailman. While I know my way around Postfix, troubleshooting Mailman is another issue.
When troubleshooting Mailman, you generally want to start with the steps outlined in FAQ 4.78 at <http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq04.078.htp>. You should also search the FAQ Wizard for other things that might be related to your problems, and you should also search the archives.
Since I don't know how to troubleshoot Mailman, I can only suspect that is has something to do with the NFS, although another part of me is telling me that it can't be.
If NFS is involved anywhere in the picture, odds of Mr. Murphy raising his nasty head have been increased by many, many orders of magnitude.
That said, there are occasionally times when such a system has problems that are actually unrelated to NFS.
Thanks for the detailed explanation. One thing to mention is that this
mail server is pushing close to 1 million e-mails a day, and has no
issues with that whatsoever. Both mail servers mount the /home for the
usernames, and everything is stored in Maildir. I'm almost positive
there are no issues with NFS itself, as all other mail works just fine.
Mailman is installed on both mail servers. /usr/lib/mailman is NOT NFS
mounted (meaning its installed on both machines), however I've NFS
mounted the /var/lib/mailman directory, where all the lists and such are.
Now, with that being said, you mentioned that Mailman can be unfriendly when placed in an NFS environment. How can I test to see that it is an NFS issue with Mailman?
I'll take a look at those docs and reply back to here. Thanks again.
Patrick
On 9/6/07, Patrick M. wrote:
Thanks for the detailed explanation. One thing to mention is that this mail server is pushing close to 1 million e-mails a day, and has no issues with that whatsoever.
I was the Sr. Internet Mail Admin for AOL. We regularly did several million mail messages per day, after filtering out 95-99% of the incoming flood as spam. One thing I learned is that, many times when you have a problem, it's with a part where you think everything is okay.
I'm not saying that I'm definite you're wrong about this, I'm just saying that you shouldn't necessarily be making any assumptions about where your problems lie, but you should instead be prepared to measure and be certain.
Both mail servers mount the /home forthe usernames, and everything is stored in Maildir. I'm almost positive there are no issues with NFS itself, as all other mail works just fine.
One line I frequently hear is "It works just fine in Windows, so obviously your Unix box must be broken". Riiiiiiiiiight.
Sorry, this is a pet peeve of mine, especially when it comes to something like Maildir.
Part of the problem is that Maildir does a number of things that are not necessary in order to solve the problem(s) it was trying to solve, and it doesn't fully solve even those issues. Meanwhile, it introduces a whole host of new issues, and this makes Maildir a much less than ideal solution for the problem.
There are other mailbox directory solutions which really do resolve the issues in question and without introducing a whole host of new problems to deal with.
Again, I'm not saying you should be forced to switch, just that you should be aware of the potential issues with your chosen solution, and you should be able to fully troubleshoot and deal with those problems.
Mailman is installed on both mail servers. /usr/lib/mailman is NOT NFS mounted (meaning its installed on both machines), however I've NFS mounted the /var/lib/mailman directory, where all the lists and such are.
This is sounding like a binary package install version of Mailman. That's not where we put things in the source version from the gnu.org/lists.org website.
In our version, you can put all of /usr/local/mailman on NFS, and as well as NFS can be made to work for anything, it should work for Mailman. This includes the Mailman queues as well as the archives, the source code, the compiled-on-demand object code, the whole kitten-and-caboodle.
Now, with that being said, you mentioned that Mailman can be unfriendly when placed in an NFS environment. How can I test to see that it is an NFS issue with Mailman?
It's not Mailman per se. It's *anything* on NFS. Mailman will work about as well on NFS as anything else can possibly work on NFS, but simply throwing NFS into the mix is likely to cause a whole host of new issues that you're going to have to deal with that other site administrators might not.
These aren't Mailman problems per se, but NFS problems that cause problems with applications running on NFS, such as Mailman.
Now, your NFS server will have a huge impact on what problems you see and what tools you have available to you to help resolve those.
Generally speaking, if you're using high-end equipment like Network Appliance, EMC, or Hitachi Data Systems, then if you have any problems you should be talking to your account representative and your support team. They're going to know much more about your hardware than we do, and they should know enough about your application to be useful.
NetApp, EMC, and HDS have gotten their stuff down to the point where Oracle actually certifies their products as being suitable for use on these servers, whereas Oracle explicitly rules out any support for their products on other NFS servers. There's a reason for this.
So, the hardware you have available to you will have a huge impact on your systems.
I'll take a look at those docs and reply back to here. Thanks again.
If you have specific questions, please let us know.
-- Brad Knowles <brad@shub-internet.org> LinkedIn Profile: <http://tinyurl.com/y8kpxu>
Patrick M.wrote:
Please forgive me, as I'm very new to Mailman. While I know my way around Postfix, troubleshooting Mailman is another issue. I can see that e-mail gets sent and delivered to the particular mailing list in question, but sometimes people tell me their e-mails never get posted to the mailing list. Other people say they submit their post, and it eventually gets there, after 10 or so minutes.
Since I don't know how to troubleshoot Mailman, I can only suspect that is has something to do with the NFS, although another part of me is telling me that it can't be.
Can you folks push me in the right direction? What should I be troubleshooting? Is there anything I can provide so that you folks can help me troubleshoot further?
Brad has given lots of good advice already in this thread including a reference to <http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq04.078.htp>. I want to emphasise that, as the thread seems to be concentrating more on NFS.
In particular, look at Mailman's 'smtp' log and correlate that with your MTA logs. Normally, you will see in your MTA log that a post has been delivered to Mailman at some time and then you will see an entry Mailman's 'smtp' log x seconds later which says "<message-id> smtp for n recips, completed in y seconds". Normally x is only slightly larger (less than 2 seconds) than y. y is the time for all the SMTP transactions to deliver the mail to the n list recipients, and y-x is the sum of the times the message was queued in Mailman's in and out queues and Mailman's non-SMTP processing time.
Also, are the 'missing' posts in the list's archive? Do the delayed posts appear in the archive without delay?
-- Mark Sapiro <msapiro@value.net> The highway is for gamblers, San Francisco Bay Area, California better use your sense - B. Dylan
participants (3)
-
Brad Knowles -
Mark Sapiro -
Patrick M.