[Spread-users] Thrudb - Document Oriented Database Services

[Alaric Snell-Pym]

On 7 Nov 2007, at 1:15 am, Jake Luciani wrote:

> Hey Alaric,
>
> Thanks for the heads up! I'll have to add the chunking.
>
> I'm using memcached knowing it forgets things BUT since writes
> happen right away and since the system can handle errors gracefully
> I think its a decent solution.

Yeah, as long as it just retries rather than exploding or anything,
that'll be fine - it'll happen so rarely that a bit of slowdown in
that case is nothing, and just retrying will ensure correctness is
preserved!

> Empirically, I've bombarded the system with gigs of data with no
> hiccups.

Talking of errors, today I've been adjusting the way mine handles
fatal errors - I originally went with the philosophy that a write
failure meant the local node was somehow broken (out of disk, I/O
error, etc), so it should shut itself down rather than continue
running and end up serving out of date data (since it couldn't
process a write).

However, the ingenious folks who write the application that sits on
top managed to trick it into broadcasting a write to a nonexistant
table, which caused every data storage node to think it was broken
and die...

Most amusing, since it was a live system!

ABS

--
Alaric Snell-Pym
Work: http://www.snell-systems.co.uk/
Play: http://www.snell-pym.org.uk/alaric/
Blog: http://www.snell-pym.org.uk/?author=4