Hi, I have a few questions related to replacing a ...
# ask-questions
a
Hi, I have a few questions related to replacing a data source in GB: 1. Our new data source will have all the historical data from the old data source, but I don’t see any easy way to transition old experiment analysis settings to the new data source, except manually going through all historical experiments and replacing the data source and adding (duplicate) metrics to the analysis settings. Is there a better way to do this? 2. Is there a way to change the data source for a metric? It’s quick and easy to just duplicate metrics attached to the old data source and create new metrics for the new data source (and this is what we have done). However, this means historical experiments are still connected to the old data source that will be decommissioned. Most metrics will be duplicates of the old metrics, so it would be much easier if it’s possible to just update the data source for a metric so it will support both new and historical experiment reports. Is this possible? 3. What will happen to result reports on historical experiments after we decommission the old data source? My understanding is that results are cached/stored even if the raw data is removed. But what happens if someone attempts to update the results? Will it remove the historical results when attempting to update, or will we still be able to load the historical results even after a failed update attempt?
f
looks like we currently don't have a way to edit just the data source for a metric - there is nothing inherently blocking this, just not exposed in the UI. I suggest that you open a GitHub issue for this request, we might be able to get to it.
for 3. if you update, and the data source fails, the query will fail. I believe it should show you the last results, but I haven't checked
are you using the cloud or self-hosting?
a
cloud
I tried to simulate 3) by breaking the data connection to a data source (from GB), but I was not allowed to update connection parameters if it breaks the connection. The other approach would be to simulate it from the data source side, but that will require more work. So if you have a quick way of testing this to know what will happen, that would be very much appreciated
Previously when we have had failed query updates on reports (e.g. when a query times out) I think it removes the results of the previous queries. Not sure, it's been a while. But if this is what happened, I'm concerned this could also happen if someone attempts to update an experiment report on a decommissioned data source