asyncd: bode transfer needs to include checkpoints

[madduck]

trafficstars

When munin contacts a host to obtain all the data asyncd as spool using spoolfetch, it seems to do so in an all-or-nothing kind of way. If the transfer gets interrupted in between, it must be restarted from the same timestamp as before.

It would be desirable that any data fetched once isn't refetched. Ideally, spoolfetch would obtain the data from one day across all plugins are once, commit those, and then move to the next day. If interrupted, the process can be resumed from the last day committed.

[Sephtex]

Hi @madduck , I would suggest to decrease the interval (1d by default) of munin-asyncd and increase the retain as I suggested in #1365 . I also gave feedback in that PR that the order of the spoolfetch files needed to be sorted in order to get this fixed.

Yet I'm not sure if your issue here is exactly the same, but after I've tested the changes that I suggested on the PR, the backlog was correctly read.