The metadata for the stats streams say to keep those events around for
864000000s which should be 10 days. After I scavenge the database via
EventStore.TestClient.exe they are still sticking around. These events
are quite large and I would like to get rid of them to improve
performance on catchup subscriptions. Am I missing something? Or do I
need to upgrade to 2.0?
Try your math again 
those are seconds. 864000000 / (60*60*24) is days.
That said in 2.0 you can change that number easily (update metadata
with $maxAge/$maxCount).
Cheers,
Greg
Yeah, as Greg pointed out 864000000 seconds = 10000 days 
Still internally $maxAge is actually set to 10 days.
Can you see in ES logs the records about starting to scavenge?
Maybe TestClient’s scavenge command doesn’t work anymore. The best way to trigger scavenging is to go to web UI to Admin page and manually trigger scavenging there.
:facepalm:
well, I was never good at math anyway I confirmed that 2.0 does scavenge the records like I expect. Will have to block out some time to upgrade then. About how long should the Projections upgrade tool take?
How big is your database?
6gb
That should be reasonably fast then.
Yep, it just completed! I have one other question before I upgrade. So
you guys took the stream metadata out for 2.0. I can see there's a new
stream 'hello' where event 0 is just a regular event. But after the
upgrade these metadata events were not removed (I understand why). I
know that the common domain example code skips event 0. Do you guys
have updated example code to handle this situation?
The old ones will still be there. However you can now update them with
the set metadata.
You no longer need to skip first event in new streams.
so in other words I will have to detect this when rehydrating an
aggregate. No problem!
Look for starts with $