The rule alerts states that "Global Rollbacks were 100% on server 'server1'. This exceeds the critical threshold of 50%.
What does this mean?
1. What do you mean by rule "Global Rollbacks were 100% on server 'server1'"?
Enterprise applications execute business functionality when transaction takes place. Websphere defines these transactions can either be "local" or "global". When a transaction is sucessfully completed, it is called as "COMMIT". But, when a transaction is failed, it is called as "ROLLBACK".
These transactions have finite amount of time to complete. When the time is exceeded, then this is called as "timeout". The rule "Global Rollbacks" assess the health of "JTA" state and nature of associated rollbacks and timeouts. When transactions fails for server1 within single JVM against more than 50% occurences then it will raise alarm for server1.
2. How this rule is calculated?
The following raw metrics for WebSphere performance is calculated as given below:
a. globalRollbackCount
b. globalCommittedCount
c. globalTimeoutCount
Then, following derived metrics is calculated:
GlobalRolledBackCounterLastInterval = (globalRollebackCount)
TotalGlobalTransactionsPerInterval = (globalCommittedCount + globalRollebackCount + globalTimeoutCount)
Therefore, percentage is calculated based on these two following arguments:
(GlobalRolledBackCounterLastInterval/TotalGlobalTransactionsPerInterval)*100
Please look at the Foglight data browser for these two metrics on their websphere server>JTA nodes and compare the data for that time.
