You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Another reason to deprioritize this is that we only have one local vat (comms), and it doesn't use a transcript anyways, so loading it back in is pretty cheap.
OTOH I don't think we've ever experienced it being evicted, so I'm not even sure we can have confidence that it would get paged back in correctly. But @dckc 's STR should exercise that pretty easily. So one task here is to just reproduce that and make sure comms still works afterwards. If that's true, we can probably put this off until we enable the creation of more vats (MN-2?)
we only have one local vat (comms), and it doesn't use a transcript
Nice! That addresses this issue, to my satisfaction.
I don't have any reason to believe that comms doesn't survive being swapped out, but if it does not, that seems like a separate issue (though I suppose morphing this one into it works for me).
Describe the bug
The LRU policy from #2848 applies to all vats, including local vats. Paging a local vat out and in is likely to be expensive and should be avoided.
As @warner notes, our defaults are probably generous enough to avoid evicting any vat, so it's unlikely to be observable any time soon.
To Reproduce
Steps to reproduce the behavior:
maxVatsOnline
to something like 7agoric start
ps ax | grep xsnap
local
vatsExpected behavior
local
vats don't get considered for swapping out (eviction).The text was updated successfully, but these errors were encountered: