Open AlinaHovakimyan opened 2 months ago
@AlinaHovakimyan - all of this is quite possible. We didn't exhaustively capture all data. Mostly we're looking at the data that helps tune gc behavior and some of these (like mark stripe, etc) aren't useful to for tuning in our opinion. We can be wrong about this and maybe you know what to do with this data. We welcome your input here.
@dsgrieve we do agree that this infos are not that commonly used, but we often see cases in low-latency applications (such as trading) where every ms of a pause matters. It's always helpful to get as granular insights as possible into why the pause takes as much as it takes to help diagnosis of what can be improved, even e.g. in the application code.
ZGC missing data: # Safepoints
DestroyDetachedPages, PrepareRelocationSet from concurrent times
mark summary (Mark: 1 stripe(s), 2 proactive flush(es), 1 terminate flush(es), 1 completion(s), 0 continuation(s) )
nMethods (registered/unregistered)
References
Metaspace: parsing exists but most probably has some bug, so comes as null
relocation info