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
Pusher's memory use pattern turns out to be bad for systems with lots of old data. It creates one tarfile for every day/datatype combination with un-uploaded data. This means that using it on a directory with lots and lots of unuploaded days causes a memory spike as all those days are loaded into independent tarfiles and uploaded independently.

I think the right thing is probably for main to be:
for each datatype {
in a goroutine {
upload all existing data serially // <-- THIS IS THE NEW PART
start listening for new data
}
}
That way, starting pusher with lots of old data won't lead to a memory spike - that data will just get uploaded serially.
The text was updated successfully, but these errors were encountered:
Pusher's memory use pattern turns out to be bad for systems with lots of old data. It creates one tarfile for every day/datatype combination with un-uploaded data. This means that using it on a directory with lots and lots of unuploaded days causes a memory spike as all those days are loaded into independent tarfiles and uploaded independently.

I think the right thing is probably for main to be:
That way, starting pusher with lots of old data won't lead to a memory spike - that data will just get uploaded serially.
The text was updated successfully, but these errors were encountered: