Exporting data between retention policies with different shard durations

Store
, ,
1

Hi,

I’m copying data between 2 influxdb servers. It’s the same database name on source/destination servers but they have different shard durations (1w on the source and 1d on the destination).
The “copying” is done by running influx_inspect export on the source and influx -import on the destination. I noticed that the import actually altered the shard duration on the destination and even worse there are shards with different durations and overlapping windows. e.g.

id database retention_policy shard_group start_time           end_time             expiry_time          owners
-- -------- ---------------- ----------- ----------           --------             -----------          ------
15 test     autogen          15          2020-07-07T00:00:00Z 2020-07-08T00:00:00Z 2020-09-06T00:00:00Z 
4  test     autogen          4           2020-07-10T00:00:00Z 2020-07-11T00:00:00Z 2020-09-09T00:00:00Z 
2  test     autogen          2           2020-07-11T00:00:00Z 2020-07-12T00:00:00Z 2020-09-10T00:00:00Z 
22 test     autogen          22          2020-07-06T00:00:00Z 2020-07-13T00:00:00Z 2020-09-11T00:00:00Z 
19 test     autogen          19          2020-07-12T00:00:00Z 2020-07-13T00:00:00Z 2020-09-11T00:00:00Z 
30 test     autogen          30          2020-07-13T00:00:00Z 2020-07-20T00:00:00Z 2020-09-18T00:00:00Z

Now i’m guessing this is probably a use case that’s not supported. But how would influxdb deal with these shards? What is the behavior when an existing RP gets altered with a different shard duration? Say you shorten the shard duration from 1w to 1d during the week.

Thanks.

2

It turned out to be a daily scheduled ansible job that resets all the retention policies, which led to the messed up shards.