I have a customer using our software, which has influx integrated into it, for storing computer performance metrics. we use a java http connection to write line feed protocol to the /write api endpoint.
The scenario we came across is that only “some” of the customers data was being inserted to the database, even though influx was providing a 204 response code back.
It was strange, because other sources of data were being inserted fine into other measurements. When we tried to insert the same “failing” data into a new measurement, the measurement wasn’t even created, even through we still got a 204 response code. The last data that was inserted, was suspiciously at 23:59:59, right at the border of midnight. which makes me think this may have been a shard creation/corruption problem. not sure. Our failing use case was inserting 600 timestamps, very small amount of data, 1 timestamp every second. However, if we inserted a much larger amount of data, the measurements, and everything seemed to work fine. As we were troubleshooting, we figured this was some kind of influx bug, and decided to upgrade from the 1.1.0 to 1.2.0, and voila, the test case worked. Infact. not only did the test case work, suddenly, in this measurement where data had stopped being inserted April 2nd, 23:59:59, there was data until April 6th, 15:30:00, as if all those “failed” writes from the past, were not failed, but finally got flushed out, or at least, that’s the only thing we could guess.
My question is this, is there a known bug that was fixed between 1.1.0 and 1.2.0 that would explain this behavior? Our customer is looking for an explanation why their data failed to be inserted.