Analysis - Default Backfill configuration for service startup
When a failover occur or at service startup, ALL analysis configured for backfilling executes based on the last configured backfill/recalculate timerange.
There is an urgent need for a DEFAULT per-analysis configuration in case of service startup, therefore not using the last executed backfill but to have it's own planned recovery which I hope would include the last executed date as an option.
ADMIN RESPONSE
Aug 20, 2022
Hi, can you please provide some details on what problems you're trying to solve?
In our case we could tackle the problem of a long time of recalculations after a startup. If we need to restart analysis Service (and this is quite regularily unfortunately), Analysis service needs between 9 and 12 hours to recalculate all analysis. As there are some analysis providing EventFrames monitored and evaluated on a daily base, the recalculation sometimes leads to a situation where the events are not finalized before evaluation takes place, leading to requests of the affected users. We try to reduce the load of automatic recalculations as good as possible, which means that we manually cancel the backfill for every database we know, that Backfill is not needed. On a Database level, this is possible to handle manually, but often you have a mixture of analysis within a database that don't neccessarily need to recalculate and analysis that definitely need to recalculate. Separating them manually could work via categories but manually cancelling them still is a lot of effort.
In our case we could tackle the problem of a long time of recalculations after a startup. If we need to restart analysis Service (and this is quite regularily unfortunately), Analysis service needs between 9 and 12 hours to recalculate all analysis. As there are some analysis providing EventFrames monitored and evaluated on a daily base, the recalculation sometimes leads to a situation where the events are not finalized before evaluation takes place, leading to requests of the affected users. We try to reduce the load of automatic recalculations as good as possible, which means that we manually cancel the backfill for every database we know, that Backfill is not needed. On a Database level, this is possible to handle manually, but often you have a mixture of analysis within a database that don't neccessarily need to recalculate and analysis that definitely need to recalculate. Separating them manually could work via categories but manually cancelling them still is a lot of effort.