Sequences for reconsiliation

 

Some important reconsiliation schemes are schetched and recovery issues are discussed.

Required information

Knowledge is assumed of TelOrb, Jambala, HLR/IN applications (mobile telecom nodes) and network redundancy.

 

 

Wake up

A zone wakes up after zone reload. The other part, stand-by or primary, is in action as a takeover or netshared.
Regardless if it is the primary or the standby zone that wakes up, it will enter down mode (update netshared data is not allowed) and request reconciliation from the other part.

 

 

 

 

 

If left zone fails, it just retries from start. Right zone should continue. If right zone fails or if contact is broken, the left zone has to do different things at different states:

EarlyNegotiation TakeOver immediately.

PrepareClientRec,

ReadyToRecClient Zone reload, in order to be available for TakeOver. Then wait for TakeOver or retry if right zone is found in takeover or netshared mode.

ReconcilingClient Continue until (A) to have a reasonable state for TakeOver.

Giveback (smooth)

GiveBack is always "smooth" and will be rolled back if the transactions counters don't match.

Both zones will at any time be ready to takeover.

Smooth takeover

Smooth takeover will be rolled back if the transactions counters don't match.

Both zones will at any time be ready to takeover.

Forced Take-Over at NetShared

The right zone receives TakeOver order.

Right zone does not care if left zone fails.

If right zone fails or if communication is broken before (A), left zone is prepared for and expects TakeOver. If right zone fails or communication breaks after (A), left zone will stop in DOWN.

Remote synch

Below describes a successful quick synch where no external synch is necessary on the left zone. The backup time sent from the left zone to the right zone is the time stamp for the backup generated during the quick synch on the right zone.

External synch

The left zone receives a new backup from right zone.

If right zone or communication fails, the Synch Done message should be repeated until ack'ed, or until newer Synch Done is performed.

If at Backlog Progress End the backlog still contains transactions that are older than the checkpoint time, probably there is a problem of version counters that have wrapped around. Left zone should go DOWN.

If there are fresh transactions in the backlog when the Backlog Progress has Ended, it is probably due to missed transactions which prompts for a new external synch - which will be performed anyhow.

If the backlog grows out of bounds - i.e. some transactions have to be completely discarded, a new reconsiliation has to be made. Again that reconsiliation will be done anyhow.

Note that left zone is consistent and available for TakeOver during the reconsiliation.