Data Race Conditions
In a QB Channel -- I seem to be experiencing a race condition From a staging table: 1. [Thread 1] System checks to see if order A exists in the sales table. Does not find Order A 2. [Thread 2] Sy...
Forum Discussion
MarkShnier__You
Qrew Legend
Qrew LegendIs the situation that you are mass creating or importing children and you want to make Parents when required? Is the parent getting duplicated? can there be a custom Key field for the parent so it can't get duplicated? ie does the trigger know what the Key field of the Parent Order should be?
------------------------------
Mark Shnier (YQC)
Quick Base Solution Provider
Your Quick Base Coach
http://QuickBaseCoach.com
mark.shnier@gmail.com
------------------------------
------------------------------
Mark Shnier (YQC)
Quick Base Solution Provider
Your Quick Base Coach
http://QuickBaseCoach.com
mark.shnier@gmail.com
------------------------------
- Hi Mark - thanks for looking at this!
A sample of data looks like this:
OrderID Date Status Product 73906 19-Mar-21 10:19:55 RESERVED 123 Hat 73906 21-Mar-21 4:08:22 COMPLETED 123 Hat
In one csv file, I'll have the order's state progression. Using a QB Record Create trigger pipeline on a staging table - writing to a Sales table.
After the file has been processed through the pipeline, I might have one record in the Sales table or two - it's seemingly random. It seems like a classic data race - https://devopedia.org/race-condition-software but I haven't been able to sort out QB's handling for it. Has anyone else seen it?
------------------------------
Malcolm McDonald
------------------------------- MarkShnier__YouYou have not really described what the pipeline is doing but I believe they do run in parallel. So for example if the pipeline is potentially triggered by every record being processed in a CSV file then to pipelines may be running simultaneously, unaware of each other.
Can you explain what the pipeline is supposed to do?
------------------------------
Mark Shnier (YQC)
Quick Base Solution Provider
Your Quick Base Coach
http://QuickBaseCoach.com
mark.shnier@gmail.com
------------------------------- Hi Mark -
I got the 'd'oh' solution from Support just now, seems to be working: require the Sales ID field on the Sales table be unique. Simple miss.
But for posterity's sake, in case anyone else is looking for this later:
I have a file of n records. The file comes from a client's system, and I have very limited opportunity to modify the file in any way at source. The file is a running history of transactions, ie I could have one record for a sale that shows a completed, fulfilled sale or I could have more than one record for a sale if it went to a pending status first, then got fulfilled or cancelled. In the file I will have multiple sales.
I needed to be able to load the file into a staging table, run some data validation steps on the records (does the product definition exist, is the sales date in the future etc), then look into the destination 'sales' table to see if I have that sale already (then update with the new information) or create the sale.
What the pipeline was doing was falling over itself; creating the update as a new sale before the original was done creating. I had a file of 30 records, 22 unique sales -- I would get anywhere between 23 and 29 records in Sale when the pipeline was complete. By checking 'Unique' on [SaleID] I now get 22 records each time.
------------------------------
Malcolm McDonald
------------------------------
