I've had similar trying to -resynch a bitcoin-qt 0.11.0 full node.
It resynchs for a couple of days, only needing two or three reboots and restarts without -resynch,
and then at seemingly random times through the past year or two worth of blocks, it hangs and the next restart won't get past "Activating Best Chain" before it hangs with "Error reading from database, shutting down".
I eventually found that it would -reindex from block 1 in 2009 but that takes days and has crashed every time eventually.
I'd be supporting the blockchain with a full node by now, a full 10 days after upgrading, if it would.
What do people think of some coding changes so that
on its own, -reindex does what it has always done; reindexing from the start
and new code can accept an =integer after -reindex
such that -reindex=280000 starts the reindexing from block 280000
and gets it a little further before it crashes again.
from a crash at 342000, you could for example
bitcoin-qt -reindex=340000
and eventually get synched and up to date, going over errors if necessary
If I see three ++ comments then I'll find a suggestion box to copy this post to. If not then its just my grotty old computer which has a problem so it is not worth bothering the devs with it.
It resynchs for a couple of days, only needing two or three reboots and restarts without -resynch,
and then at seemingly random times through the past year or two worth of blocks, it hangs and the next restart won't get past "Activating Best Chain" before it hangs with "Error reading from database, shutting down".
I eventually found that it would -reindex from block 1 in 2009 but that takes days and has crashed every time eventually.
I'd be supporting the blockchain with a full node by now, a full 10 days after upgrading, if it would.
What do people think of some coding changes so that
on its own, -reindex does what it has always done; reindexing from the start
and new code can accept an =integer after -reindex
such that -reindex=280000 starts the reindexing from block 280000
and gets it a little further before it crashes again.
from a crash at 342000, you could for example
bitcoin-qt -reindex=340000
and eventually get synched and up to date, going over errors if necessary
If I see three ++ comments then I'll find a suggestion box to copy this post to. If not then its just my grotty old computer which has a problem so it is not worth bothering the devs with it.
i am against your idea for two reasons:
1) reindex isnt done regulary. so it doesnt really matter if it takes a while.
2) if someone enters the wrong blockindex for his reindex it may lead to very strange and hard to debug errors which may even go unnoticed for a longer time. better stay safe
not sure if it is even possible - this would only work if the index is handled by bitcoin-core and not by bdb - and i dont know how its implemented.
