Anyone who's looked at the stats lately has probably wondered a couple of
things. First, where all the new blocks are coming from and second, how on
earth we could still be doing blocks yesterday...
Well, the stats are a bit wonky right now. The bad new is, I will not be
able to repair things until about 14:00 GMT. Here's what happened, for
those who are curious.
Here's how the stats *should* be processed:
1. Logfile is automagically transferred from the master keyserver
to the stats server
2. Stats server "notices" the new logfile, and begins processing
3. New logfile is processed into the database and made live on the
web pages.
4. Logfile is moved to archives for backup and storage.
5. Stats server waits for another logfile to appear.
However, the machine that handles "backup and storage" has gone casters-up
for some reason (which is odd because it's normally a very stable machine).
So, when step 4 was to occur, the Logfile never actually moved anywhere.
Then, when Steps 1 and 2 took place again the next hour, the old, stale,
already processed logfile was re-processed and re-posted. Again and again,
the duplicate data getting larger and larger with each new processing event.
Oops.
All for lack of a little error testing in my code. Slap me on the wrist.
I've halted all the processing, so the stats will be static until I am able
to fix things. The only way to really be certain that things are right
will be to completely re-build yesterday and today from the logfiles. I
won't be able to do this until the dead box is brought back online.
I apologize for any confusion or inconvenience this may have caused.
Oh, and yes, the code now checks to make sure that the file was
successfully moved out of the active directory before continuing now.
-/\/ugget
---- To unsubscribe, email majordomo@llamas.net with 'unsubscribe rc5-announce' in the body.
This archive was generated by hypermail 2b30 : Fri Sep 26 2003 - 05:53:00 EDT