Poor man's failover... - Solaris

This is a discussion on Poor man's failover... - Solaris ; I'm looking to do a POC of a poor man's failover and was curious if anyone knew of any existing open source software and/or scripts that will handle this... Essentially, I will have two servers with six zones (including the ...

+ Reply to Thread
Results 1 to 8 of 8

Thread: Poor man's failover...

  1. Poor man's failover...

    I'm looking to do a POC of a poor man's failover and was curious if
    anyone knew of any existing open source software and/or scripts that
    will handle this...

    Essentially, I will have two servers with six zones (including the
    global). Here is what I am thinking...

    First server:
    Five public IP's for all zones except the database zone, which will
    have a private network that the zones will communicate with on loopback
    (192.168.15.x/24).

    Second server:
    One public IP (to access to the Global Zone), then the other five zones
    will be on the private network that will be heart-beat monitored
    through a cross-over cable on a second interface. The database will
    run on both servers, first server is the master, the second server is
    the slave and running replication with MySQL. Other zones have pretty
    much static data and will be rsync'd on a schedule that makes sense.

    Here's the theory....
    The heart-beat will monitor the first server, if an outage occures, it
    will then re-IP itself to the other servers public IP, change the
    database from slave to a master and *poof* work like magic.

    Thoughts/Input?


  2. Re: Poor man's failover...

    On Sat, 22 Sep 2006, te@ivorypetal.com wrote:

    > Here's the theory....
    > The heart-beat will monitor the first server, if an outage occures, it
    > will then re-IP itself to the other servers public IP, change the
    > database from slave to a master and *poof* work like magic.


    One product which springs to mind is RSF-1 from High Availability

    http://www.high-availability.com/links/2-8-rsf1.php

    It's very simple and is primarily script based (using S and K scripts).
    In the event of a primary failure the secondary detects the lost
    ethernet/serial/disk heartbeats and the scripts start up all the
    relevant services. RSF-1 takes care of MAC address and storage
    paths/device names so that they retain the same identity. It's then
    down to the applications to do their thing to recover - eg, rollback
    from the transaction log.

    --
    Chris

  3. Re: Poor man's failover...

    On 22 Sep 2006 21:07:36 -0700, te@ivorypetal.com wrote:

    >I'm looking to do a POC of a poor man's failover and was curious if
    >anyone knew of any existing open source software and/or scripts that
    >will handle this...


    For simple things linke IP failover, Linux-ha (http://linux-ha.org/)
    should work. But attention, documentation is very poor.

  4. Re: Poor man's failover...

    In alt.solaris.x86 fabrice-pas-despame.bacchella@worldonline.fr wrote:
    > On 22 Sep 2006 21:07:36 -0700, te@ivorypetal.com wrote:
    >
    >>I'm looking to do a POC of a poor man's failover and was curious if
    >>anyone knew of any existing open source software and/or scripts that
    >>will handle this...

    >
    > For simple things linke IP failover, Linux-ha (http://linux-ha.org/)
    > should work. But attention, documentation is very poor.


    Also, despite what it may say on the web site, getting Linux HA working on
    solaris is a struggle.

    --
    Geoff Lane, Airstrip One


  5. Re: Poor man's failover...

    On 23 Sep 2006 19:20:14 GMT, wrote:

    >Also, despite what it may say on the web site, getting Linux HA working on
    >solaris is a struggle.


    Yes, indeed, as getting Linux HA working on anything is already a
    struggle.

  6. Re: Poor man's failover...

    #!/bin/ksh
    while [ 1 -gt 0 ]
    do
    ping 192.168.1.190 > /dev/null 2>&1
    if [ $? -ne 0 ] ; then
    ifconfig dmfe0:1 192.168.1.190 netmask 255.255.255.0 > /dev/null 2>&1
    ifconfig dmfe0:1 up > /dev/null 2>&1
    ...may need to restart/start app here....
    ...maybe email your pager to alert you
    exit 0
    fi
    done

    Nohup the script so when you logout it doesn't go away. If the host has
    "lom" interfaces, you can even echo power off sequences over the serial
    cables as to be sure the other host is actually powered off.

    There will be those that argue the above is not perfect, it isn't. But has
    major advantages in that it is simple, trivial, not a whole lot to
    understand and read. And cetianly don't need a specialist and/or 5 days of
    reading to pull off.

    Not being overly complex has value too.


    wrote in message
    news:1158984456.179261.285980@m73g2000cwd.googlegr oups.com...
    > I'm looking to do a POC of a poor man's failover and was curious if
    > anyone knew of any existing open source software and/or scripts that
    > will handle this...
    >
    > Essentially, I will have two servers with six zones (including the
    > global). Here is what I am thinking...
    >
    > First server:
    > Five public IP's for all zones except the database zone, which will
    > have a private network that the zones will communicate with on loopback
    > (192.168.15.x/24).
    >
    > Second server:
    > One public IP (to access to the Global Zone), then the other five zones
    > will be on the private network that will be heart-beat monitored
    > through a cross-over cable on a second interface. The database will
    > run on both servers, first server is the master, the second server is
    > the slave and running replication with MySQL. Other zones have pretty
    > much static data and will be rsync'd on a schedule that makes sense.
    >
    > Here's the theory....
    > The heart-beat will monitor the first server, if an outage occures, it
    > will then re-IP itself to the other servers public IP, change the
    > database from slave to a master and *poof* work like magic.
    >
    > Thoughts/Input?
    >




  7. Re: Poor man's failover...

    On 22 Sep 2006 21:07:36 -0700 te@ivorypetal.com wrote:
    > The heart-beat will monitor the first server, if an outage occures, it
    > will then re-IP itself to the other servers public IP, change the
    > database from slave to a master and *poof* work like magic.
    >
    > Thoughts/Input?


    Pretty much what I do. I use ucarp for heartbeat.

    -frank

  8. Re: Poor man's failover...

    This is simple to achieve, but you will have to do some work here.

    01. Between 2 zones, you will have to have 3 separate ip-addresses on
    each server.
    example: 1 in 172.x network and the other 2 in 192.x network.

    02. Write scripts on both servers to ping each other on private network
    of 172.x network. Here 192.x network is assumed to talk to the rest of
    the world ( meaning to the global zone).
    02.a. If a ping on 172.x fails for more than 5 seconds, then ping the
    public ip-address.
    02.b. If this public ip-address fails, then ping or check the port
    usage of the service you are running on the other server. If that too
    seems to be down, then go to step 3

    03. Bring up the interface with and assign the hostname.

    04. Have a script running on the Global zone which checks for a file's
    contents to change from 0 to 1

    05. Once the service is up, change a preset file content to 1. This
    will be picked up by the global zone and you can have it do what you
    want, viz, email / page / make sure zone-1 is down etc.

    These scripts either has to run via crontab or from inittab.

    Let us know how it goes. Good luck.

    -GGR
    Rajiv G Gunja


+ Reply to Thread