Bug in monthly collection, days 1-31

[huitema]

There was a bug in the monthly collection data of "by day" data. The line was:

    if (ret) {
        this->hourly_volume[time_m.tm_hour] += 1;
        this->daily_volume[time_m.tm_mday] += 1;
    }

This is wrong, because tm_mday has values from 1 to 31. On day = 31, the buffer overflowed. It is not clear whether this affected the next record in the data file:

    uint64_t query_volume;
    uint64_t hourly_volume[24];
    uint64_t daily_volume[31];
    uint64_t arpa_count;

The values of arpa_count seem correct, but this is worth an investigation.

The record is shifted by 1. We can fix that in the pandas analysis by changing the heads of the column, restating the first column as "d00", and computing "d31" as queries - sum(d01..d30).

Then, with pandas, we can compare the new d31 to the counter arpa. If we find a high correlation, this becomes suspicious!