Page 2 of 2 FirstFirst 12
Results 11 to 16 of 16

Thread: Zimbra on Pacemaker + DRBD howto

  1. #11
    adrian.gibanel.btactic is offline Senior Member
    Join Date
    Nov 2010
    Posts
    54
    Rep Power
    4

    Talking

    Quote Originally Posted by Taer View Post
    Hi Folks!

    Just wanted to say that I tried the bTactic Zimbra OCF script with Zimbra 8 OSE & Ubuntu 12.04 x64, and all is working fine.
    That's a great piece of news.
    Quote Originally Posted by Taer View Post
    If it can help, this is how I integrated the Zimbra script (invented nothing, that's what the spanish doc is saying) :

    On each VM of the cluster, do the following :

    Code:
    mkdir /usr/lib/ocf/resource.d/btactic
    - Put the script in this folder.

    Code:
    chmod +x /usr/lib/ocf/resource.d/btactic/*
    On one of the VMs enter:
    Code:
    crm configure
    , and inject the config :
    Code:
    primitive ZMServer ocf:btactic:zimbra \
    op monitor interval="120s" timeout="40s" \
    op start interval="0" timeout="360s" \
    op stop interval="0" timeout="360s"
    Thank you! These small pieces of code where you explain how you have done things yourself are very meaningful for people that try to understand how all this stuff it goes.

    And, of course, it clarifies that the script works for someone else than the original writer.


    Quote Originally Posted by Taer View Post
    Thank you for this wonderfull tool Adrian. It made me save a lot of time.
    You're welcome. If you ever document your work don't hesitate to share a link here.

    It would be nice to know if the zimbra mysql gets corrupted somehow when one of the active/pasive cluster machines dies. This first version of the howto does not replicate mysql as a normal mysql replication but with files thanks to DRBD as you might know.
    Adrian Gibanel, IT Manager

    "be free, be innovative, bTactic"

    Av. Balmes, 34 4rt 1a B | Lleida, Spain 25006

    www.btactic.com

  2. #12
    adrian.gibanel.btactic is offline Senior Member
    Join Date
    Nov 2010
    Posts
    54
    Rep Power
    4

    Post English translation to come

    As a part of my University thesis is highly probable that this howto is:

    • Translated into English
    • Improved to be accepted as a University thesis paper
    • to be added mysql replication (not sure about that)
    • Adapted (if needed) to Ubuntu 12.04 and Zimbra 8
    • Use more common network setups (not sure about that)

    .

    Just in case anyone wanted to translate the whole howto into English. I will do myself in the next six months.
    Last edited by adrian.gibanel.btactic; 10-29-2012 at 03:58 AM. Reason: Minor writing fixes
    Adrian Gibanel, IT Manager

    "be free, be innovative, bTactic"

    Av. Balmes, 34 4rt 1a B | Lleida, Spain 25006

    www.btactic.com

  3. #13
    Taer is offline Junior Member
    Join Date
    Oct 2012
    Posts
    6
    Rep Power
    2

    Default

    Quote Originally Posted by adrian.gibanel.btactic View Post
    That's a great piece of news.


    Thank you! These small pieces of code where you explain how you have done things yourself are very meaningful for people that try to understand how all this stuff it goes.

    And, of course, it clarifies that the script works for someone else than the original writer.
    You're welcome, I'm glad to help.

    Quote Originally Posted by adrian.gibanel.btactic View Post
    If you ever document your work don't hesitate to share a link here.
    I am indeed curently writing a modus operandi for the archives of the IT Department, when it'll be finished I will be happy to translate it in english for sharing with the community.

    Quote Originally Posted by adrian.gibanel.btactic View Post
    It would be nice to know if the zimbra mysql gets corrupted somehow when one of the active/pasive cluster machines dies. This first version of the howto does not replicate mysql as a normal mysql replication but with files thanks to DRBD as you might know.
    So far I got no problems with a single dead machine. I started reading and working about HA something like 2 weeks ago, but if I understood correctly the working of DRBD it keeps all the modifications in memory so when the other node is alive and ready for sync again, it sends them. And until now, this was confirmed by all my tests. But I'll try to lost my Master during a serie of zmprov operations. (Wouldn't be surprised if Zimbra's devs put a COMMIT in all LDAP provisioning tools thought.)
    And I'm a little concerned about split brain issues too, anyway I'll keep you update of that all.


    Ho, and good luck for your University thesis.

  4. #14
    Taer is offline Junior Member
    Join Date
    Oct 2012
    Posts
    6
    Rep Power
    2

    Default

    Hello everyone!

    Got some news today, I made a little how-to explaining how I built my Zimbra cluster:

    https://www.dropbox.com/s/8w4f7koz9y...HA_Zimbra8.pdf

    A few words about it, to catch on:
    This how-to explains the HA installation of Zimbra 8 on an Ubuntu server 12.04 LTS x64, the HA will be made between two VM named fangorn (preferred master) and
    sylvebarbe (slave).
    Two different storage technics were tried: DRBD (for all Zimbra files: /opt/zimbra), and virtual bus sharing by VMware on a SAN (for index and mails, which are by default on /opt/zimbra/index and /opt/zimbra/mail).

    And I made some tests, all my results are included on the above PDF, I already can say here that I haven't noticed any corruption of Zimbra's database. But it would be nice if someone else could try and confirm... or not.

  5. #15
    adrian.gibanel.btactic is offline Senior Member
    Join Date
    Nov 2010
    Posts
    54
    Rep Power
    4

    Default

    Congratulations!

    It seems I missed your message. Zimbra 8 HA in Ubuntu 12.04 seems not as different as my original howto finally.
    I'm going to update my howto as a University thesis paper anyways, probably in 6 months time.
    Your howto will help me a lot.

    Adrián Gibanel
    Adrian Gibanel, IT Manager

    "be free, be innovative, bTactic"

    Av. Balmes, 34 4rt 1a B | Lleida, Spain 25006

    www.btactic.com

  6. #16
    adrian.gibanel.btactic is offline Senior Member
    Join Date
    Nov 2010
    Posts
    54
    Rep Power
    4

    Post Zimbra 8 High Availability on Ubuntu 12.04 Master Thesis

    Here there is my thesis about Zimbra 8 high availability: Zimbra 8 High Availability on Ubuntu 12.04 master thesis and its associated presentation: Zimbra 8 High Availability on Ubuntu 12.04 master thesis presentation .

    Some comments about it that are not explained in thesis itself:

    • The thesis does not take care of periodic crontab running on the passive node (/var/spool/crontab/zimbra)
    • You can also use other shared storages other than DRBD but that's not documented
    • Split DNS is not explained in the guide. If Split DNS is needed configuring Bind on the Zimbra server will have to be performed on both nodes. Bind configuration should be the same one on both servers. You can keep them in sync manually. Anyways Bind should run in its own and should not be managed by Pacemaker at all. Well, you can make Pacemaker manage it for saving some RAM resources but I don't think it pays the effort.
    • Probably an image explaining how DRBD RAID1 network based partition synchronisation works should be added


    If someone wants to implement its own Zimbra High Availability system you can use my own thesis as a guide.

    Some features from the thesis:

    • The different cluster components are depicted as Virtualbox virtual machines so that you can experiment your HA system in your own Virtualbox (or favourite virtualization software) without having to deal with physical machines till you are ready.
    • Operating system installation is detailed explaining the DRBD-Metadata partition
    • Each of one of the technologies involved is explained just before using them so that you understand them
    • Zimbra installation steps are detailed compared to other guides which want you to know how to install Zimbra. Although it's recommended you had already done it, of course.


    As I have already mentioned before if you have any doubt about Pacemaker configuration I recommend you to read Clusters from Scratch document which you can find at master thesis' bibliography so that you can understand how everything works in detail.

    I want to thank Zimbra community for their previous work on HA, professors from Universitat de Lleida, and, of course, bTactic crew, the place where I work.

    Last edited by adrian.gibanel.btactic; 11-02-2013 at 11:46 AM. Reason: Added link to master thesis source code in order to be able to download OCF file
    Adrian Gibanel, IT Manager

    "be free, be innovative, bTactic"

    Av. Balmes, 34 4rt 1a B | Lleida, Spain 25006

    www.btactic.com

Page 2 of 2 FirstFirst 12

Thread Information

Users Browsing this Thread

There are currently 2 users browsing this thread. (0 members and 2 guests)

Similar Threads

  1. Zimbra not playing nice with Pacemaker
    By BCamp in forum Installation
    Replies: 6
    Last Post: 08-04-2013, 12:45 PM
  2. Zimbra+Pacemaker need help
    By belajarlinux in forum Administrators
    Replies: 1
    Last Post: 04-26-2012, 02:03 PM
  3. Zimbra + DRBD need advice
    By Samp in forum Administrators
    Replies: 1
    Last Post: 09-14-2010, 03:35 PM
  4. [SOLVED] CentOS 5.4 with LVM & DRDB/Pacemaker?
    By ekkas in forum Administrators
    Replies: 0
    Last Post: 03-17-2010, 04:47 AM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •