Got some indexed events that were incorrectly timestamped, like set to 20 years into the feature, and would like to know if the DELETE command would help in removing these bad events from the index.
Concern with these bad events, if not removed:
1. bucket will not roll to frozen/archiving. Currently, the frozen policy is 1-year.
2. performance impact, by having these events around
The delete command will not remove data from the index properly, it will just exclude the events from future searches.
To avoid these wildly wrong timestamps from getting into your index you could look at reducing your MAX_DAYS_HENCE setting in props.conf, by default it's just a couple of days. Anything much larger rarely makes sense.
As for rolling the affected bucket to frozen, it might indeed not roll it on its own because it contains extremely new events far from the future. You could manually roll the bucket though, once you're certain the other correctly timestamped events are due for a freeze.
Warm and cold buckets' directory names contain the newest and oldest event timestamp in the bucket - in your very specific case of events indexed 20 years into the future that's enough to find the buckets containing badly timestamped events.
Another lazy way would be to wait a year for the natural roll to frozen, and pick out those that should have rolled but didn't.
Is there a way to determine which buckets are affected by the bad events?