Looking closely over the ingestion code, it seems there is at least one noticeable issue: one year after we remove the suite level value, all the tests in their own right (and which are children of the suite signature) will be deleted with all their associated perf data points. Why? If a suite no longer outputs a value, no data points are ingested (it doesn't have what to ingest). In time (one year), our expiring algorithms will delete all its existing data points. When a signature no longer has data points, it is itself deleted; even if it's a parent signature. As children signature are configured to be deleted when their parent signature is deleted (in a CASCADE fashion), we end up removing still active data.
Bug 1672794 Comment 2 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
Looking closer over the ingestion code, it seems there is at least one noticeable issue: the subtest signatures get decoupled. New signatures will be created for them. As we already know, decoupling perf data leaves our perf sheriffs blind for at least 24h. Given that we run our jobs more rarely now, this blind period could be higher, up to a couple of days.
Looking closer over the ingestion code, it seems there is at least one noticeable issue: the subtest signatures get decoupled. New signatures will be created for them, which will collected new data points separately. All old signatures (both parent & child) will eventually be deleted (in approximately a year). As we already know, decoupling perf data leaves our perf sheriffs blind for at least 24h. Given that we run our jobs more rarely now, this blind period could be higher, up to a couple of days.
Looking closer over the ingestion code, it seems there is at least one noticeable issue: the subtest signatures get decoupled. New signatures will be created for them, which will collect new data points separately. All old signatures (both parent & child) will eventually be deleted (in approximately a year). As we already know, decoupling perf data leaves our perf sheriffs blind for at least 24h. Given that we run our jobs more rarely now, this blind period could be higher, up to a couple of days.