Saturday, May 16, 2009

New simultaneous viewers record and related slowness

Seems like we hit a new high of over 1000 simultaneous map viewers today, mostly thanks to Dayton Hamfest, and a popular live Hamfest video feed with an embedded map. They're giving away freebies.

A couple of components started hitting file descriptor limits, which had last been upgraded over a year ago. Too many simultaneous connections per process. This made the site perform very, very slowly. I quadrupled the limits, and the site started to perform quickly again, I hope that's enough for more than a year to come. Well, I have to admit that it would actually be a nice surprise if the site would be so popular that it wouldn't be enough...

I also fixed a bug in the "first heard" algorithm pointed out by Ian, VK1IAN. The digipeater alias GATE was not treated as a special digipeater alias (like WIDE, RELAY and TRACE are), and an igate which first heard a packet with a GATE in the digi path was not given credit for hearing it first.

Another fix that went in was a filter which takes out complete APRS packets which have somehow made their way to the comment field of another APRS packet. Apparently something is loosing CR LF sequences between packets (could be my code...), which causes packets to go into the comment of the previous one. Before I find the actual bug I've added a filter to strip these off.


Mark (KB9KHM) said...

It looks like there may be a new bug in that was just introduced recently. It appears that when looking at a D-Star station by entering their callsign in the "Track callsign:" box, or opening with the ?call= parameter, it shows the digi path for all D-Star stations has having gone through a digi in Germany.

This issue does not occur if you click "clear" and view all of the stations in a given area, then the D-Star stations are shown with proper digi paths.

Niclas Cronsioe said...
This comment has been removed by the author.
oh7lzb said...

Mark: I'll look into it. I have an idea why it happens.