|
Home > Archive > EAserver > April 2005 > Failover support
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
|
|
|
| EAS 5.2
PB 10
Hi !
I have this configuration: A cluster with 2 jaguars and a pb
component (stateless) with "automatic failover" and
"transaction required" using a connection cache.
I need to do a test shuting down one jaguar using the pb
component. The idea it is that the other jaguar executes the
component, and this is correct but I need to be sure that
"the transaction is rolled back" because if don=b4t the
component could repeat the transaction"
Some body could confirm that point...
Please, help....
| |
| Dave Wolf 2005-04-28, 11:24 am |
| Yuri wrote:
> EAS 5.2
> PB 10
>
> Hi !
>
> I have this configuration: A cluster with 2 jaguars and a pb
> component (stateless) with "automatic failover" and
> "transaction required" using a connection cache.
>
> I need to do a test shuting down one jaguar using the pb
> component. The idea it is that the other jaguar executes the
> component, and this is correct but I need to be sure that
> "the transaction is rolled back" because if donīt the
> component could repeat the transaction"
>
> Some body could confirm that point...
>
> Please, help....
1) What is your client?
2) Failover does NOT insure consistency. There can be cases (very well
times cases) where data modifications could occur twice.
Imagine this scenario.
EAS1
EAS2
PB component in both does an insert.
Client calls component on EAS1. The component activates, connects to
the db, inserts the row, commits the work, deactivates, and is
juuuuuuussssst about to marshall success to the caller, when boom EAS
goes down.
Now, the client sees this as a server down failure, he client side ORB
executes a failover, and calls the failover server.
Client calls component on EAS2. The component activates, connects to
the db, inserts the row, BOOM! Work is done twice.
I call this "Double Dipping".
Guess what? This isnt EAS specific. It affects every HA app server on
earth.
Now, you can test failover, but there is nothing the container can do
about double dipping. You need to design around that. Now to be fair,
the odds of it are low... but it can happen.
Dave Wolf
Cynergy Systems
http://www.cynergysystems.com
| |
| Chris Pollach 2005-04-28, 1:27 pm |
| But very good if you are making a large Bank Deposit to your savings
account <lol>!
"Dave Wolf" <dave.wolf@cynergysystems.com_nospam> wrote in message
news:4271110b$1@foru
ms-2-dub...
> Yuri wrote:
>
>
> 1) What is your client?
> 2) Failover does NOT insure consistency. There can be cases (very well
> times cases) where data modifications could occur twice.
>
> Imagine this scenario.
>
> EAS1
> EAS2
>
> PB component in both does an insert.
>
> Client calls component on EAS1. The component activates, connects to
> the db, inserts the row, commits the work, deactivates, and is
> juuuuuuussssst about to marshall success to the caller, when boom EAS
> goes down.
>
> Now, the client sees this as a server down failure, he client side ORB
> executes a failover, and calls the failover server.
>
> Client calls component on EAS2. The component activates, connects to
> the db, inserts the row, BOOM! Work is done twice.
>
> I call this "Double Dipping".
>
> Guess what? This isnt EAS specific. It affects every HA app server on
> earth.
>
>
>
> Now, you can test failover, but there is nothing the container can do
> about double dipping. You need to design around that. Now to be fair,
> the odds of it are low... but it can happen.
>
> Dave Wolf
> Cynergy Systems
> http://www.cynergysystems.com
>
>
| |
| Dave Wolf 2005-04-28, 1:27 pm |
| Unless you're the bank ;)
Dave Wolf
Cynergy Systems
http://www.cynergysystems.com
Chris Pollach wrote:
> But very good if you are making a large Bank Deposit to your savings
> account <lol>!
>
> "Dave Wolf" <dave.wolf@cynergysystems.com_nospam> wrote in message
> news:4271110b$1@foru
ms-2-dub...
>
>
>
>
|
|
|
|
|