Welcome to Qlik Community forum and thanks for reaching out here!
TSWP Files are internal binary format files which store records temporary, Qlik Replicate usually keeps transaction data in memory until it is fully committed to the source and/or target. However, transactions that are larger than the allocated memory or that are not committed within the specified time limit will be offloaded to disk TSWP Files.
Qlik Replicate will try to read the change records from TSWP Files to memory and apply them to target side automatically if possible. You need NOT care about the TSWP Files unless Replicate failed to process these files at all. There is no way to operate these TSWP Files manually directly.
Hope this helps.
John.
Help users find answers! Do not forget to mark a solution that worked for you! If already marked, give it a thumbs up!
In this scenario, am I correct in understanding that any data exceeding the thresholds specified in "Transaction Offload Tuning" will be written to TSWP files?
Also, could you please clarify when these TSWP files are automatically deleted?
Hi, Yes, so for better performance set it to be longer in duration than the batch setting. Also, the Total transactions memory size exceeds (MB) setting in Transaction Offload Tuning will come into play too. In normal circumstances TSWP files are automatically deleted once they have been loaded back into memory for processing. If you run into Task issues they could remain. Cheers!
Thank you.
Is my understanding correct that TSWP files are generated under either of the following conditions?
① When the "Transactions duration exceeds (seconds)" setting in "Transaction Offload Tuning" is set to a lower value than the "Longer than (seconds)" setting in "Batch Tuning."
② When the volume of updates exceeds the "Total transactions memory size exceeds (MB)" setting in "Transaction Offload Tuning."
Regarding the point that TSWP files are normally supposed to be deleted upon being committed to the target but sometimes remain undeleted:
In my testing, it appears that they are rarely deleted at the first "Longer than (seconds)" interval;
instead, I frequently observe two of them being deleted together at the second "Longer than (seconds)" interval.
YW. Yes that is correct. Think of it as a temporary transaction store in case a Task is showing high latency and transactions start hitting a backlog. You don't want this backlog consuming all of the memory on the Replicate server since no Task would be able to run in that scenario.
I guess there could be some timing mismatches on what you're observing. Is this observation affecting something?
We aim to reduce Snowflake warehouse costs by setting the application interval to once per hour.
However, extending the interval resulted in the creation of TSWP files, making it necessary to re-evaluate the Qlik server resources.
If TSWP files were deleted upon application to the target, we would only have needed enough capacity for one hour's worth of data; however, since there were instances where they were not deleted, we wanted to clarify the system's specifications regarding this behavior.
Hi, It sounds like you need a more detailed look at what is happening in terms of loading the swap files back into memory and then deleting them in your environment, and maybe then some Task optimisation, so that would be Professional Services. Or if you think the Task is not behaving as expected then Support may be able to help. Or maybe an easier route is to provision a bigger disk as part of the resource eval. Cheers!
