My understanding is that this feature prevents the truncation of active transaction logs by executing transactions against an internal table (attrep_truncation_safeguard) in the source database while unread logs are being processed/read.
Based on this understanding, could you please clarify the following two questions?
1. Release Timing of Internal Transactions At which of the following timings (①–③) is the internal transaction issued by this feature released? If it is none of these, please let us know the exact timing.
① Upon completion of log reading via fn_dblog ② Upon completion of log parsing/filtering within Qlik Replicate (identifying changes that need to be propagated to the target) ③ Upon completion of data propagation to the target database
2. Methods for Explicitly Releasing Internal Transactions Is our understanding correct that stopping all tasks using this source database as an endpoint is a valid method to explicitly release these internal transactions? If there are any other methods to explicitly release them, please let us know.
The safeguard transaction is used to protect unread TLOG records during log capture/read. It is not held until data is applied to the target. The closest option is one (log reading/capture stage), not three.
2. Explicit release:-
Yes. Stopping all Qlik Replicate tasks that use the source endpoint and have "Start transactions in the database" enabled will release the safeguard transactions.
Below are some documents that gives you detailed information about the same.
