Linux HA solution

[ust3]

I have previously post wrong forum so I post to here again,

I have two live linux sever , they run different application , have different IP address & different user a/c , both of them connect to the SAN server with Fibra channel , the hardware has already been setup and works fine , the SAN has clustering function that mean the database is physically located at SAN (different path) , both servers can access all these database . We have windows DNS server to resolve the host name , so user can connect the specific server with the host name .

Now , I would like to setup the HA solution , what I would like is when one of server is broken then another server can take up the responsibility , allow user to login .

My question is when a server is broken , how it can automatically re-direct the user from broken server to another server ? Do I need to setup LDAP ? Do I need to do something in DNS server ?

Can anyone advise what I need to do

Thx in advance.

[Ivan Ferreira · Accepted solution]

The best solution is to use a combination of technologies. What distribution are you using?

Red Hat Cluster Suite is my favorite HA solution for clustering.

Yes, you should maintain accurate user accounts using LDAP.

>> The SAN has clustering function that mean the database is physically located at SAN (different path) , both servers can access all these database

Don't start confusing concepts. The Storage may provide shared access, but this is not clustering. Shared file system (cluster file system) is required for simultaneous access. This is provided by the O.S. for example, with Oracle OCFS2 or Red Hat GFS.

[Colin Topliss]

You can look at using ServiceGuard, or LifeKeeper - they will provide that for you. However, what you'll end up with is users accessing their application via a Virtual IP Address (VIP). In the case of ServiceGuard, each VIP is associated with its relevant application (which is under ServiceGuard control).

Take a look at:

http://h18026.www1.hp.com/solutions/enterprise/highavailability/linux/serviceguard/index.html
http://www.steeleye.com/products/linux/

There are of course other possible solutions (as mentioned by the previous poster) - some are cool, some are a bit clunky. It depends on what you are after!

Regards

Colin.

[Jimmy Vance]

As others have said, HP Serviceguard, or SteelEye LifeKeeper are your best bets.

LifeKeeper is a bit more intuitive to use than Serviceguard.

Serviceguard and LifeKeeper are similiar as they use virtual IP's. LifeKeeper can failover either of your applications to the other system when it fails, I'm pretty sure Serviceguard will do the same.

[Steven E. Protter]

Shalom,

Logins can easily be handled in a number of ways.

Samba integration to a foreign Windows ADS server.

NIS

LDAP

LDAP is the preferred choice. Master/slave setup.

The rest can be handled by Red Hat Cluster suite.

A DNS based solution will not work well because DNS is not really smart enough to know whether the target node is responding. Just point the DNS A record to the floating IP address of the cluster.

SEP

[Serviceguard for Linux]

To follow up on Jimmy's comment - Serviceguard can do the same. Applications are monitored as well as the HW. If an application fails, you can set up the cluster to try to restart it some number of time before failing over if you wish. If the HW fails then all applications are failed over.

IP failover works the about same for most clusters. So as was mentioned, it is probably better to set up an IP address and a server alias name for the application rather than trying to have the DNS know when to fail over.

[Pierguido]

One solution could be also heartbeat.
I'm using it right now on a testing environment and it works pretty well.
You'll have a VIP (virtual IP) to connect to the cluster and heartbeat is taking care to do all the work in the case one server fails.
You can check on http://www.linux-ha.org/

Pier