We have discovered a fault in networking interface if the stations are not synchronized.
The result is that we can have 2 valid records with the same timestamp for a variable.
It is possible to reproduce it systematically with this architecture:
2 servers in real time association (single active server) and in historical association (multiple active server).
The passive station is one minute late comparing to the active station.
The association is archiving a counter that is incremented by 1 every minute.
The trend group is configured at 1 minute.
If you switch the passive station to be actif and wait 1 minute you will have 2 records with the same timestamp.
If the trend group is configured at 5 minutes it is more difficult to reproduce because the switch must be done just after the trend group period as explained in the SPR62866.
It is supposed to be fixed in version 12 and the referent guys for Networking and HDS will take care of it in order to have a more robust and stable version of PcVue.
Until then, the good practice is to maintain the stations synchronized and to use deadband for counters.
The bad idea is to remove the Value field of the table in SQL from the primary key. It would considerably reduce the performance when PcVue try to read the archives.
This behavior is particularly bad for DreamReport counter statistical function because the result of such functions is totally wrong if we archive 2 values with teh same timestamp.


