multi server - running with slave while master is down - postfix queries master ldap

[su_A_ve]

Hello,

We currently have a multi server setup. Primary server has all services enabled, secondary server has all services except logger enabled and ldap is set as a replica.

If we need to bring down the Primary server to patch OS for example, we were under the impression that secondary would still work for their own local mail store. Login to the GUI works fine, however we cannot send out email.

Looking in the logs, it seems postfix is trying to query the master ldap.

Nov 10 15:46:03 xxxx postfix/trivial-rewrite[17296]: warning: dict_ldap_lookup: Search error -1: Can't contact LDAP server
Nov 10 15:46:03 xxxx last message repeated 2 times
Nov 10 15:46:03 xxxx postfix/trivial-rewrite[17296]: fatal: ldap:/opt/zimbra/conf/ldap-vad.cf(0,100): table lookup problem
Nov 10 15:46:04 xxxx postfix/smtpd[10554]: warning: premature end-of-input on private/rewrite socket while reading input attribute name
Nov 10 15:46:04 xxxx postfix/smtpd[10554]: warning: problem talking to service rewrite: Success
Nov 10 15:46:04 xxxx postfix/master[19566]: warning: process /opt/zimbra/postfix-2.2.9/libexec/trivial-rewrite pid 17296 exit status 1
Nov 10 15:46:04 xxxx postfix/master[19566]: warning: /opt/zimbra/postfix-2.2.9/libexec/trivial-rewrite: bad command startup -- throttling
Nov 10 15:46:04 xxxx postfix/smtpd[10014]: warning: premature end-of-input on private/rewrite socket while reading input attribute name
Nov 10 15:46:04 xxxx postfix/smtpd[10014]: warning: problem talking to service rewrite: Connection reset by peer

ldap-vad.cf still points to the master ldap.

[bdial]

don't you have to promote it to master?

[su_A_ve]

Why would I have to? My assumption is that while the master is down, I could still operate on those accounts under the 2nd server.

It seems postfix is set to contact the master only. On the 2nd server (ldap replica) ldap_master_url is correctly set, ldap_is_master is set to false, and ldap_url is correctly configured to point to itself first and then the master, however the ldap-***.cf files only point to the master.

[LMStone]

Originally Posted by su_A_ve

Why would I have to? My assumption is that while the master is down, I could still operate on those accounts under the 2nd server.

It seems postfix is set to contact the master only. On the 2nd server (ldap replica) ldap_master_url is correctly set, ldap_is_master is set to false, and ldap_url is correctly configured to point to itself first and then the master, however the ldap-***.cf files only point to the master.

Don't have the specific command handy, but wanted to reply to you fast...

In the Multi-Server Install Guide, there is a zmWhateverICantRememberNow command that is a local, per-server config setting to tell the local server in what order it should query LDAP servers in the farm.

When we set up an additional mailbox/mta server as an LDAP replica in a Zimbra farm, we run this command to change the order in which that server will query LDAP servers.

For each mailbox/MTA server that is also an LDAP replica, we set the order to be:

localhost, OtherLDAPReplica1, OtherLDAPReplica2, LDAPMaster

In this way, the LDAP Master is not taxed, and each server makes LDAP queries at bus, rather than network, speeds. Plus, if you configure all your mailbox/MTA servers this way, when one server is down it won't impact the other servers.

Hope that helps,
Mark