We’ve discovered that foreground SQL latency can be affected by changefeed backfills (initial scans and catchup scans). Let’s see if we can reduce the size of this struct.
@nvanbenschoten mentions that it might be a small as key, current value and previous value -- down to 3 pointers instead of the current 10.
Jira issue: CRDB-17821
Epic CRDB-23739
We’ve discovered that foreground SQL latency can be affected by changefeed backfills (initial scans and catchup scans). Let’s see if we can reduce the size of this struct.
@nvanbenschoten mentions that it might be a small as key, current value and previous value -- down to 3 pointers instead of the current 10.
Jira issue: CRDB-17821
Epic CRDB-23739