OC4J 18.104.22.168.0 (ejb2.0), Oracle9i, Swing client for a three tier setup...
- Problem has gotten worse over time, with code that have not been altered
- Problem persists in persistence(pun intented), we experience double insertions of data, for no apparent reason
- Business logic that should prevent this kind of behavior is not present or not executed
- User presses button once, 99% of the time only one set of inserts happen, sometimes two (few oocassions more, one time 130!!), swing client will act like 'hanging'
- Timestamp comparrision in one case revealed that there were 5 seconds between the dupes (equals 5 second client hang?)
hypothesis ; connections/operations succeed but return a false state, thus the application layer(JDBC driver?) repeats the task ? Other than that, a far fetched idea about a flawed connection pool in OC4J for transaction handling (wich does not explain the entire scenario) i am kinda blank on this one..
Of course the problem is non repeatable in 'safe envoriment' ... only happens in wild (read : production !!!!!)
Have ANYONE seens something like this before ? Im on the verge of blaming the hardware here .. nic's routers whatever ... i cant see what else ..
Funny thing is, that this HAS worked flawless for like 6mo's .. then suddenly... and in decreasing intervals...