Ah yes, the block writer being overqueued on spinning disks is a
problem, especially since we're indexing the full history and other
data. Writes will be optimised further:
https://wiki.unsystem.net/index.php/Libbitcoin/Blockchain

On 04/01/14 13:37, William Swanson wrote:
> On Sat, Jan 4, 2014 at 3:46 AM, Robert Williamson <bobalot@???> wrote:
>> This has happened to me a few times before, I think it is due to how
>> libbitcoin handles the initial download of blocks while it is catching up to
>> the current height. I think we're requesting getblocks from too many peers
>> at once, or timing out too quickly (a peer might be under load or have a
>> slow/congested connection) and we're getting another copy of the data later
>> on, leading to the duplicate objects.
>>
>> This usually isn't a problem, leave it for a while and it will eventually
>> restart the download and catch up, maybe we need to add something for
>> handling the initial block download when we know we're more than a few
>> thousand blocks behind, even after the bootstrap, there is still a large
>> amount of data that we need to get.
>
> Ah, this makes sense. Another interpretation might be that the block
> writer thread can't keep up, and that it's work queue is overflowing.
> I know for a fact that the block writer is IO-bound on my hard disk,
> and the blocks are coming in awfully fast. This would also explain why
> things are fine until about 1000 blocks have been downloaded, and then
> start getting progressively worse.
>
> Either way, I'll just follow your advice and let things be for now.
>
> -William
> _______________________________________________
> Libbitcoin mailing list
> Libbitcoin@???
> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/libbitcoin
>