And in the meantime, we can live with a 1 minute latency.

Vendors estimated they can deliver the permanent fix by mid January, but I'm so fucking relieved we finally figured out what was wrong, and I can go back to sleep my usual minimum 8 hours a day.

I'm tired boss.

Vendor from Source A has a sleep timer of 1 second, not allowing all the records to be properly read by our B layer -the ETL platform-

We figured out a way to apply a property that avoids the sleep timer completely, and now the latency is down to less than a minute.

We are still far from the 5 seconds SLA, but still, much, much better than those fucking 30 minutes.

Proper fix has to be developed by vendors A and B, as both have wrong code not suitable for this particular data pattern

All this time we thought we screwed it up somehow. Data latency was 30 minutes. 30 fuck-ing mi-nu-tes.

Data's SLA for this particular application is 5 seconds, as this is a real-time software built for one of the most important areas of the company core business.

There was some crazy retention happening in the Source of the data, and this week we discovered the root cause of the problem...