Terracotta Discussion Forums (LEGACY READ-ONLY ARCHIVE)

Hi !!

I tested tc servers in multicast environment a month ago. It looks work well. But I found something strange when I tested.

First question.
My web application didn't reponse if both tc servers downed. I want my web application still work even though both tc servers downed.

Second question.
My web application still didn't work after I re-started both tc servers again. My web application only works after restart it.
Is it normal ?

Any comment will be helpful !!

Thanks

msbaek71 wrote:

Hi !!

I tested tc servers in multicast environment a month ago. It looks work well. But I found something strange when I tested.

First question.
My web application didn't reponse if both tc servers downed. I want my web application still work even though both tc servers downed.

Terracotta isn't designed to survive a double failure at the server layer (any # of clients can fail and your cluster will continue working).

Double failures are extremely rare, and almost never designed to be survived. For example, typical web applications that rely on a database do not continue to function in the face of a complete failure of the database layer.

That said, the 2.3 version of Terracotta introduced Cluster Membership events that are sent via JMX. You can trap a JMX event that indicates that the Terracotta server is unreachable, and respond to it however you like.

msbaek71 wrote:

Second question.
My web application still didn't work after I re-started both tc servers again. My web application only works after restart it.
Is it normal ?

No, this is not normal. Please try the latest version of Terracotta 2.4 and ensure your configuration is correct. If you still have the problem post your configuration and we can try to sort it out.

Note that even in the Network Active Passive mode, if you want to survive restarts (in your case, the restart of both nodes) then you need to enable Persistent mode.

We have a new Terracotta Server Cluster doc that describes setting up the Terracotta Server Cluster here: http://www.terracotta.org/confluence/display/docs1/Terracotta+Server+Cluster

(Note that it is brand new, so your comments are welcome to improve it)

Terracotta isn't designed to survive a double failure at the server layer (any # of clients can fail and your cluster will continue working).

I am not sure if this was already clear, but you can have ANY number of passive (standby) Terracotta servers as you wish (it doesn't have to be just 2). For example, you could deploy n number of Terracotta servers behind a dual redundant network to survive a network switch failure.

Regards,
Kunal Bhasin.

Also worth noting that if the active and passive are run in persistent mode then either could just be restarted and things can continue as normal.

kbhasin wrote:

Terracotta isn't designed to survive a double failure at the server layer (any # of clients can fail and your cluster will continue working).

I am not sure if this was already clear, but you can have ANY number of passive (standby) Terracotta servers as you wish (it doesn't have to be just 2). For example, you could deploy n number of Terracotta servers behind a dual redundant network to survive a network switch failure.

Regards,
Kunal Bhasin.

Kunal - yes that's right - I was thinking in just the context of the original post. I should have said Terracotta isn't designed to survive a complete failure. Since as you point out there can be more than 2 servers, this can have as many backups as one would like, and all have to fail. In the context of the original post, that is a double failure.

And as Steve notes, if all you want is to bring them back, that's easily done by enabling persistent mode.