All messages deferred - lost connection with 127.0.0.1 RCPT TO

[bill@hurricanelabs.com]

I sent this off to support but thought I might post here in hopes of someone just knowing the solution (I'm pretty sure I know the issue). Anything with an x was just edited to remove domain specific information. Thanks in advance:

We're attempting to migrate (not upgrade) to 5.0.5 (zcs-NETWORK-5.0.5_GA_2201.UBUNTU6.20080417004026) on an entirely new machine and we're having the following issue (see log message below). All mail is being deferred. I tracked the problem down to Amavisd on port 10024 not being able to look up legitimate names from LDAP. I've no clue how to fix this.


Apr 28 09:56:09 concordex postfix/smtp[8548]: AA6175020AC4: to=, orig_to=, relay=127.0.0.1[127.0.0.1]:10024, delay=74699, delays=74698/0.06/0.21/0.06, dsn=4.4.2, status=deferred (lost connection with 127.0.0.1[127.0.0.1] while sending RCPT TO)

Debug snippet from Amavisd:

Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) TempDir:repare: creating directory /opt/zimbra/data/amavisd/tmp/amavis-20080428T104952-24382
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) TempDir:repare_file: creating file /opt/zimbra/data/amavisd/tmp/amavis-20080428T104952-24382/email.txt
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) TempDir:repare_file: layers: unix,perlio
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) lookup_ip_acl (mynetworks): key="172.19.60.10" matches "172.16.0.0/12", result=1
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) lookup (debug_sender) => undef, "billford@x.com" does not match
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) ESMTP> 250 2.1.0 Sender OK
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) switch_to_client_time 480 s, smtp response sent
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) idle_proc, 6: was busy, 7.0 ms, total idle 0.025 s, busy 0.029 s
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) idle_proc, 5: was idle, 0.2 ms, total idle 0.025 s, busy 0.029 s
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) ESMTP< RCPT TO: ORCPT=rfc822;billford@x.com\r\n
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) switch_to_my_time 480 s, SMTP RCPT received
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) lookup_acl(billford@x.com) matches key ".", result=1
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) lookup (local_domains) => true, "billford@x.com" matches, result="1", matching_key="."
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) query_keys: billfordx.com, billford, @x.com, @.x.com, @.com, @.
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) lookup_ldap "billford@x.com", query keys: "billford@x.com", "billford", "@x.com", "@.x.com", "@.com", "@.", base: , filter: (&(objectClass=amavisAccount)(mail=%m))
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) ldap begin_work
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) Connecting to LDAP server
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) connect_to_ldap: trying ldap://concordex:389
Apr 28 10:49:52 concordex amavisd/sbin/amavisd[24382]: (24382-01) connect_to_ldap: connected to ldap://concordex:389
amavisd (ch1-24382-01): symbol lookup error: /opt/zimbra/openssl-0.9.8g/lib/libssl.so.0.9.8: undefined symbol: EVP_idea_cbc

I believe the last line there is the culprit. I think the ssl libs are just linked wrong.


I've verified DNS, hostname, etc. I've verified that the ldap server answers queries correctly. Any thoughts? This worked when we were on an earlier release of 5.0 but we upgraded right before our migration because we wanted to be on the latest release. Any help before I rebuild would be appreciated. Thanks!

[uxbod]

When you say migrate are your source and target versions of ZCS the same ? How did you perform the migration ?

[bill@hurricanelabs.com]

What we did was just copy over our settings from one server to the other (first one is 4.5, second one is 5.0.5), we were then using imapsync to move the mail. I've never had good luck with any of the other methods of migration. We're moving from CentOS to Ubuntu so a straight move would just not work.

[uxbod]

Hmmm, personally I would have installed the same version of ZCS using the dummy installed, removed it, copied the old /opt/zimbra across, ran the fixperms and then upgraded.

[bill@hurricanelabs.com]

Originally Posted by uxbod

Hmmm, personally I would have installed the same version of ZCS using the dummy installed, removed it, copied the old /opt/zimbra across, ran the fixperms and then upgraded.

That method has simply never worked for me (I know it's worked for lots of other people) even with the same platform. I'm pretty sure that wouldn't work across two different platforms (CentOS vs Ubuntu).

[mmorse]

Originally Posted by bill@hurricanelabs.com

That method has simply never worked for me (I know it's worked for lots of other people) even with the same platform.

Originally Posted by bill@hurricanelabs.com

I'm pretty sure that wouldn't work across two different platforms (CentOS vs Ubuntu).

It does, the 'magic' is the ./install -s in step 3: /blog/archives/2007/10/moving_zcs_to_another_server.html
-Key is using the same ZCS version # (you can switch platforms though) another thing to note is the GID & UID's.

[bill@hurricanelabs.com]

Originally Posted by mmorse

It does, the 'magic' is the ./install -s in step 3: /blog/archives/2007/10/moving_zcs_to_another_server.html
-Key is using the same ZCS version # (you can switch platforms though) another thing to note is the GID & UID's.

All I have is my experience to go on -- I've tried about 5 times or so and it fails in various spectacular ways. I will try again, mostly because I'm a glutton for punishment. :-) Thanks.

[uxbod]

Moving ZCS to New Server - Zimbra :: Wiki

[bill@hurricanelabs.com]

Originally Posted by uxbod

Moving ZCS to New Server - Zimbra :: Wiki

We'll give these couple of methods another shot. I'm still confused as to why it didn't work with an entirely new set up but we'll see how this goes. Thanks for the help.