[RC5] Query on recent submission losses
fiddles at distributed.net
Wed Oct 20 02:22:28 EDT 2004
What you're seeing could be something as simple as a delay in the blocks reaching the keymaster. Remember, work units aren't credited to stats until they get back to the keymaster, which could be a few hours later.
On 19-Oct-2004, Per Hallstr?m wrote:
> I have the same problem, but have seen more like 1-2% slippage. Two
> days ago the pproxy sent 94,335 units, of which 93,132 were credited
> yesterday (1,203 work units lost). I thought that it was due to random
> blocks, so I imported the packet submission log (data for the past year
> and then some) into a database and made some queries, but found no
> random blocks submitted the last months. Which was what I expected,
> since the clients should have a steady supply of fresh work units at
> the pproxy.
> I need to make sure that the clients does not submit duplicate work
> units (why should they?). Have to download the logfile and import it
> into my database again. Weekend fun, maybe!
> On Oct 18, 2004, at 22:29, Jonathan Smith wrote:
> > I've been monitoring the difference between the number of blocks
> > submitted by my pproxy to distributed.net and the number of blocks
> > that are credited to me by distributed.net, call it my slippage.?
> > Since the start of October I've seen a sharp increase in the
> > slippage rate between submitted blocks and credited blocks and was
> > wondering if anyone else has noticed the same thing.
> > To provide some numbers, in September which was a fairly typical
> > month in terms of slippage I averaged:
> > ?875 blocks a day received by my pproxy server
> > ?867 blocks a day credited by distributed.net
> > for an average of 9 blocks a day slippage, or 0.87% slippage.
> > So far in October I've averaged:
> > ?869 blocks a day received by my pproxy server
> > ?775 blocks a day credited by distributed.net
> > for an average of 94 blocks a day slippage, or 10.87% slippage!?
> > Although this is an average, it is a consistent problem and not
> > the result of one or two days of exceptionally high slippage.?
> > There have been 10 days, out of 17 processed so far this month,
> > with 85 blocks or more of slippage and only one day where more
> > blocks were credited than sent (single day negative slippages are
> > normally a fairly common occurrence due to pproxy and keyproxy
> > delays)
> > Basically I'm wondering if this is an isolated problem I'm having
> > or a if it is systematic issue with distributed.net,
> > ??? Jonathan Smith???
> > _______________________________________________
> > rc5 mailing list
> > rc5 at lists.distributed.net
> > http://lists.distributed.net/mailman/listinfo/rc5
> rc5 mailing list
> rc5 at lists.distributed.net
fiddles at distributed.net
More information about the rc5