Master Master replication questions and problems

Discussion in 'HOWTO-Related Questions' started by marchost, Sep 19, 2008.

  1. marchost

    marchost New Member HowtoForge Supporter

    I set up a master master replication as explained by falko how-to.

    It works but I ran in some problems.

    Let say there is a replication between server1 and server2 (master-master)

    1)If server1 reboot, will it resync automatically with server2 upon restart?

    From what Ive seen, it doesnt seem to resync automatically but maybe I did something wrong.

    Everytime I reboot a machine I have to :

    stop slave;
    reset slave;
    reset master;
    start slave;

    because log position is not the same...

    on both machines and then find tables that are not in sync with maatkit :

    #mk-table-checksum h=server1,u=root,p=pass h=server2,u=root,p=pass | mk-checksum-filter

    and resync tables - in this example table1 on db1 (on server1 that rebooted):

    mk-table-sync --synctomaster --execute h=localhost,u=root,p=pass,D=db1,t=table1

    2)My second question is about replication in general. Again I will suppose that server1 rebooted.

    If I do (on server1)

    show slave status;

    I will get a "Read_Master_Log_Pos:" that does not match the output of "Position" :

    (on server 2)
    show master status;

    Will the replication still work??? I mean, is it necessary to do stop slave, reset master, reset slave, start slave on both to restart the replication?

    Any help would be greatly appreciated

  2. falko

    falko Super Moderator ISPConfig Developer

    Yes, provided that the server's downtime is < expire_logs_days. If not, then there are no logs so that the server doesn't know what to replicate.

    What is important is that Slave_IO_Running and Slave_SQL_Running both show Yes.
    The servers might not always be 100% in sync - depending on network load or server load it can take a few seconds until a server catches up.

    If you need to repair replication, this tutorial might be interesting:
  3. marchost

    marchost New Member HowtoForge Supporter


    Thanks for your answer Falko...

    I successfully build a load balanced + high availability server (currently in production) on 2 real servers running Xen (instead of having 4 servers...) with http, mysql, mail and dns based on many of your howtos...

    Im almost done writing an howto about it, will post it on howtoforge when it will be completed but you can take a peak :

  4. falko

    falko Super Moderator ISPConfig Developer

    Looks great! I'd love to see it on HowtoForge. :)
  5. marchost

    marchost New Member HowtoForge Supporter


    Just put up the howto on howtoforge last night (took me 3-4h formatting the tutorial from my website :( )...

    Soon will start working on a better version with shared high availability NFS for mail & web files instead of rsync, more user friendly interface (DNS, mail config, etc) and vmware.

    Its totally unrelated but i benchmarked the xen kernel with unixbench on 3 computers and I always get 50% of the score compared with the standard kernel. Do you think vmware would give better performance? Maybe unixbench is not a good real life benchmark?

  6. falko

    falko Super Moderator ISPConfig Developer

    Thanks a lot, will publish it soon! :)

    Maybe with VMware ESX, but not with VMware Server. You might also want to try OpenVZ.

Share This Page