Ceph

When PGs are created, the initial scrub timestamp is 0.000000 where it should be the current time. With the current behavior, a scrub will be immediately performed on cluster creation even though there is no data present. Because utime_t uses unsigned 32 bit integers, it doesn't appear to be possible to stop this initial scrub by setting a high interval value.

With high PG counts it can take a significant amount of time for the cluster to finish the scrubbing process. Given that it is not immediately obvious that a scrub is in progress, this could lead to users seeing lower than expected benchmarks when testing on freshly created clusters.