Posted by In the previous articles, i’ve been explain how to install and configure Zimbra on CentOS 6 or CentOS 7, how to install and configure online failover/failback on CentOS 6 using Heartbeat and how to install and configure data replication on CentOS 6 using DRBD. All above guidance could be combined to get Zimbra High Availability. For online failover/failback, you could using Heartbeat. For data replication, you could using DRBD. Heartbeat + DRBD will produce High Availability (HA). The following is guidance to configure Zimbra HA
Step by step to configure Zimbra HA
For the Linux systems, i am using CentOS 6 64 Bit. For easy understanding, this is my information system
# Server 1 Hostname : node1 Domain : imanudin.net IP Address : 192.168.80.91 # Server 2 Hostname : node2 Domain : imanudin.net IP Address : 192.168.80.92 # Alias IP Hostname : mail Domain : imanudin.net IP Address : 192.168.80.93
Alias IP will be used for access clients/users. This alias IP will be configured online failover
# install Zimbra on CentOS 6 on all nodes (node1 and node2) as described at this link : How To Install Zimbra 8.6 on CentOS 6. Please note some information below
– Please change name of each nodes refers into mail.imanudin.net when installing Zimbra
– Set IP address of each nodes refers into mail.imanudin.net include DNS and /etc/hosts
# Stop Zimbra and DNS services on all nodes (node1 and node2)
su - zimbra -c "zmcontrol stop" service named stop chkconfig zimbra off chkconfig named off
# After installed Zimbra, install and configure Heartbeat on all nodes (node1 and node2) as described at this link : How To Configure Online Failover/Failback on CentOS 6 Using Heartbeat
# After installed Heartbeat and online failover/failback working fine, then install DRBD for data replication on all nodes (node1 and node2) as described at this link : How To Configure Data Replication/Synchronize on CentOS 6 Using DRBD
# Testing data replication DRBD that has been worked : Testing Data Replication/Synchronize on DRBD
# After DRBD has been worked, copy file/folder /opt/zimbra into DRBD devices.
Do the following command only at node1
– Rysnc Zimbra
drbdadm primary r0 mount /dev/drbd0 /mnt/tmp rsync -avP --exclude=data.mdb /opt/ /mnt/tmp
data.mdb will be huge if copied by rsync so that take a long time. For the tricks, use cp for copy data.mdb to DRBD devices 😀
– Copy data.mdb
cp /opt/zimbra/data/ldap/mdb/db/data.mdb /mnt/tmp/zimbra/data/ldap/mdb/db/data.mdb chown zimbra.zimbra /mnt/tmp/zimbra/data/ldap/mdb/db/data.mdb
# Umount DRBD devices after rsync file/folder Zimbra at node1
umount /dev/drbd0
# Move folder /opt existing to another folder, do the following command on all nodes (node1 and node2)
mv /opt /backupopt mkdir /opt
# Configure /etc/hosts and dns records on all nodes (node1 and node2)
vi /etc/hosts
so that like below
127.0.0.1 localhost 192.168.80.91 node1.imanudin.net node1 192.168.80.92 node2.imanudin.net node2 192.168.80.93 mail.imanudin.net mail
vi /var/named/db.imanudin.net
change IP address of mail so that refers into IP 192.168.80.93. See the following example
$TTL 1D
@ IN SOA ns1.imanudin.net. root.imanudin.net. (
0 ; serial
1D ; refresh
1H ; retry
1W ; expire
3H ) ; minimum
@ IN NS ns1.imanudin.net.
@ IN MX 0 mail.imanudin.net.
ns1 IN A 192.168.80.91
mail IN A 192.168.80.93
# Configure file /etc/ha.d/haresources on all nodes (node1 and node2)
vi /etc/ha.d/haresources
so that like below
node1.imanudin.net IPaddr::192.168.80.93/24/eth0:0 drbddisk::r0 Filesystem::/dev/drbd0::/opt::ext3 named zimbra
# Stop service Heartbeat on node2 and then node1
service heartbeat stop
# Start service Heartbeat on node1 and then node2
service heartbeat start
TESTING HA
– Failover
After Zimbra running well on node1, please stop service Heartbeat on node1 or force off machine
service heartbeat stop
All services that taken over by Heartbeat will automatically stopped and taken over by node2. How long node2 can start all services worked again, depends how long start services (named and zimbra)
– Failback
Please start again service Heartbeat on node1 or power on machine
service heartbeat start
All running services on node2 will automatically stopped and taken over by node1
Hooray, finally you could build Zimbra HA with DRBD+Heartbeat
For log information about HA process, you can see at /var/log/ha-log
Good luck and hopefully useful 😀
Halo mas imanudin, nama saya Dekri, mau bertanya :
– Jika server nya berbeda data center dan pastinya berbeda ip address, itu bagaimana ya mas?
– Selain menggunakan Heartbeat, pacemaker dan DRBD apakah ada software HA lainya ya mas?
Terimakasih
Hi mas Dekri,
– Bisa komunikasi langsung menggunakan IP public mas. Untuk DRBD ataupun untuk Heartbeat nya
– Dibagian Heartbeat, bisa diganti ucast nya menjadi IP public lawannya
– Untuk software HA yang lain, saya belum tahu lagi mas
Is there no hearbeat package on centos 7
no package available
Hi Gillier,
Yes, you can use Pacemaker
halo, sya alvian, saat sya lakukan mount /dev/drbd0 /opt, direktori zimbranya malah hilang dan sya tidak bisa mengakses zimbranya, salahnya dimana ya…., trimaksih….
Hi mas Alvian,
Folder nya tidak hilang. Melainkan diganti dengan device drbd yang di mounting. Coba drbd nya mount dulu ke folder lain dan salin seluruh isi /opt/zimbra ke folder hasil mounting DRBD. Bisa dilihat pada bagian ini “– Rysnc Zimbra”
halo mas iman,, waktu penginstalan zimbranya apakah di node ke dua install seperti biasa atau dummy instalation ya_./install.sh -s), dan waktu mengintsall apakah file /optnya perlu di mount dlu k DRBD
thnx….
need help for my homework :))
Hi mas,
Bisa hanya jalankan ./install.sh -s saja
Hai mas aku mau nanya,, aku cluster dan penginstalan sudah selesai, tapi dalam beberapa menit clusternya down dengan log seperti ini
ResourceManager(default)[93104]: 2019/09/06_03:18:00 ERROR: Return code 1 from /etc/init.d/zimbra
ResourceManager(default)[93104]: 2019/09/06_03:18:00 CRIT: Giving up resources due to failure of zimbra
ResourceManager(default)[93104]: 2019/09/06_03:18:00 info: Releasing resource group: vn-zmb-cltr-01-uph IPaddr::10.12.1.161/24/ens160:0 drbddisk::r0 Filesystem::/dev/drbd0::/opt::ext3 zimbra
ResourceManager(default)[93104]: 2019/09/06_03:18:00 info: Running /etc/init.d/zimbra stop
Mungkin bsa dbantu mas??
Hi mas,
Coba disable SELinux nya jika pakai CentOS
Sudah dsable dri awal mas…
Hi,
I am using Zimbra mail server with 250 accounts. and I have forwarding of emails accounts to a single account.
now user didn’t get mails timely due to queues of mails on single account. around 2000 mail stuck in ques.
please help me out on this. or give solution.
Hello,
Maybe the mailbox users is busy. You can try to increase LMTP thread : https://wiki.zimbra.com/wiki/Performance_Tuning_Guidelines_for_Large_Deployments#LMTP
Hey Imanudin, how are you doing ?
I have one issue while setting up zimbra HA cluster.
Please find the scenario below.
I have one server in cloud and another in inhouse. Now what i am going to configure is , setting up two servers with same version of zimbra installation and DRBD setup. Can you please tell me how can i setup virtual ip for inhouse and cloud server ?
Thanks
Hello,
1. For your case, you do not need to configure virtual IP
2. You can create single name on DNS refer to 2 IP public. 1 inhouse IP and 1 IP on the cloud
3. Or you can create load balancer with haproxy and save in the cloud.
do you mean if we got 2 IP public, how to configure ip set on haresources? what i mean is what ip do i need to set there
Hi Erwin Alifiansyah,
You can exclude IP on haresources. Only Bind and Zimbra services that handled by haresources
Hi
After i move /opt to /backupopt. how am i suppose to start zimbra again?
“service zimbra start
warning: cannot change directory to /opt/zimbra: No such file or directory”
Hi Jonas,
The Zimbra service will automatically be started by Heartbeat
Thank you for the response Ahmed. i Noticed after i replaced keepalived with heartbeat 🙂 (heartbeat is end of life in Rhel7 / centos 7 )
If i want to change something in zimbra do i have to unmount it again and move to opt og are there a way to access su – zimbra
Hi Jonas,
You can change where :
1. DBRD act as Primary
2. Zimbra Folder has been mounted
Hello Iman,
Thanks for work you’ve done for us 🙂
do you have similar guides (HA, failover) for Zimbra 8.8.15 on Centos 7? =P
Hi Temo,
No for this time :). I still use Heartbeat right now 🙂
HI Iman,
I had some queries related to PTR records, all the mail servers requires rDNS reords ( PTR ) so in this case if iam creating a server in this private ip internal environment how this will work?
Hi Vishal,
If you sending email to outside, email will known from public IP. Even use private IP. How to know your public IP, you can check with this command
will it work on Centos 8 and Zimbra 9?
Hi Temo,
DRBD works. However, Heartbeat has been discontinued
Thanks Iman.
Use keepalived and conntrackd to replace heartbeat or use pacemaker if that’s an option for you to get virtually any system in HA, you can customize the primary-backup.sh form conntrackd to trigger any extra actions you might need while monitoring your services with 0 downtime on failover.
Thank you for this blog. helped me alot.
Hi,
Let me know if this method works on Zimbra 9 and Zimbra 10 latest versions installed on the Ubuntu 20 servers.
Also, let me know what things are needed to get a Zimbra server with High availability in a production environment. No official documentation is available on this.
I posted these questions on the Zimbra support panel and they responded with two words “Not Possible”.
1. Possibilities to upgrade mail servers from Zimbra 9 to Zimbra 10 without downtime.
2. Possibilities for applying Zimbra patch installations on production mail servers without downtime.
3. Possibilities of High Availability in Zimbra.
Hello Mallu,
1. Is not possible (upgrade inplace)
2. Is not possible. After perform patching, sometime need restart for apply patch
3. Possible. You can use infrastructure method like VMware HA, Proxmox HA, or DRBD+Heartbeat/Pacemaker