Open R-omk opened 1 week ago
I would also like to note that snowflake id values are only comparable between values of the same type of snowflake id; they cannot be compared with an arbitrary timestamp.
... especially considering that a custom time offset is used for snowflakeid generator.
Yes I agree it should be zero. You get that it you use the python utilities to dump the data. It is not an issue if the _version is always increasing. It should always be the case.
.. now there is a timestamp ms plus a serial number, but this value may be greater than the value calculated during cdc event processing, especially when the
snowflake.id
is disabled.in the example on the screenshot the version is calculated from GTID position, it is obvious that the ReplacingMergeTree engine cannot works properly
context: version: 2.3.0
snowflake.id: "false"