View Single Post
  #5 (permalink)  
Old 08-11-2006, 05:46 PM
spiderman spiderman is offline
New Member
 
Posts: 3
Default

I unfortunately didn't have time to mod the mboxes. I wish I had know sooner. Oh, well.

So, here's how it went, off the top of my head:

At 1200, I changed the IP on the old mail server, ran /etc/netstart, and changed it to some extra external IP we had. I then set up the networking on the zimbra box, and moved its cable over to the DMZ.

When it prompted me for the domain with no MX address, I put in company.com instead of mail.company.com. It worked as expected.

I set up the admin account as something other than admin, set the password, and changed all the ports.

It wasn't entirely working, so I had to killall rpc.statd and portmap, then I ran update-rc.d portmap remove to make sure it didn't come back to haunt us. I also set up a zimbra script in /etc/init.d to start up in case of reboot.

I zmcontrol stop'd twice, killed off any leftover processes, and zmcontrol started.

Then the mail queue monitor and some other things didn't work. I seached the forums and found that it does expect ssh to run on port 22, so I had to deal with that. The server runs chkrootkit out of cron regularly, just in case something does happen.

After that, it was sending everything into the deferred queue, so looking through the logs, there were what I guessed were permissions issue. Just to be sure, I ran:

chown -R zimbra:zimbra /opt/zimbra

Then after googling for the postfix errors, I found:

postfix -c /opt/zimbra/postfix/conf set-permissions
(or something like that for the conf directory)

After that and restarting, everything worked like a charm.

We had three other people to help, fortunately. My training on the setup probably wasn't good enough, as it was problematic at first. Others volunteered, but found it way too difficult to account for all these weird possibilities. My written procedures probably would require being rewritten, since the situations were so oddball at times in practice, though I tried to account for it as much as possible.

My script ran into tons of consequences, as well -- for example, some people were off-domain and had their usernames capitalized, so I had to run the script from cmd by:

set username=lowercaseusername
\\server\folder\zw.bat

And I had to continuously hack at it because of the inconsistencies and a few errors, though I at least had a decent template to do so. But it did set up their mozilla/thunderbird profiles quite nicely, and often we just had to make a single change or two. It was worth it just for the 3-4 people per person it did work automatically for.

Message filters could be copied pretty decently from a msgfilter.dat thing, too, with a bit of small post-setup. I missed that at first. And I also had to account for copying bookmarks.html to the new profile, which I forgot.

We were moving folders over for about 5 hours, and probably got half of the people in it over. Some people locked their offices and shut down their computers, even though we told them not to. We plan on doing the rest on Monday, and until then, recommending the web interface until they can be set up and have their mail copied. We've also set an experimental policy that new employees will only use the web interface.

On a few of them, we decided not to move the mail over yet, since they had tons of mail in tons of folders that would be a pain to move. We decided to leave them be where they were, and either copy them later, or just have an old network folder for a few people.

Case in point, the server actually quit receiving transferred mails on other computers when we copied over 20,000 mails (no exaggeration), some of which were possibly close to 10 years old, from someone who insisted on keeping them.

Still, we've probably copied gigs of mailboxes, and it hasn't taken up a gig yet in extra storage space. When stuff's not copying, it's fast, too, though some of that can be attributed to my scripted profile caching it by default on the hard drive in mozilla's offline folders.

Monday's going to be hell to pay, but it's all worth it. Once it gets going, it's solid and no longer a total mess. I touched base with the users while I was upgrading some of them, and they were really excited about centralized e-mail when I showed it off, and especially getting e-mail from home by web access. I gave out post-it notes to the webmail url, even, and told them to give it a try.

Sure, centralized webmail's standard for most businesses, but it's neat to see it being launched as a new thing somewhere.

All in all, it went pretty well for being a nightmare scenario.

If anyone thinks I should write up this in detail and post it here or maybe even on the wiki, because of the scenario, let me know.
Reply With Quote