Sometimes Stops Updating Shared Trackers

[morgant]

Ron Parker and I have noticed some issues with our trackers (which share with each other) sometimes losing and/or failing to update their shared listings. He writes, “It seems npdscmd.txt gets stale and I’ve had to reload my server twice. Not sure if it’s @npds or the tracker.”

[QuadzillaNET]

At about 1100 server hits after two weeks, the newtzilla npds-tracker kept registering my Newton NPDS as "Down" though it was up. Then it would disappear from my npds-tracker all-together. I tried disconnecting and re-up'ing from within the Newton Tracker Client but the results would be the same. Briefly as "Down" and then it would disappear from the tracker. At about that time, I also disappeared from UNNA's tracker. I unloaded/loaded my npds-tracker process and everything went back to normal except I never did show as "Up (Sharing)" on UNNA ever again (until @morgant reloaded the UNNA tracker).

It might take a couple of weeks to show itself again. My NPDS Newton tends to re-boot itself about every 4 hours, with all un-essential extensions frozen. When we start to (not) see ourselves on the other's tracker, let's get a dump of the npdscmd.txt file and see what's in it.

[QuadzillaNET]

Restarted my tracker today. Notice that it stopped recognizing misato.chuma.org as a shared tracker yesterday. I was still shared with www.unna.org and those servers were listed on my tracker. It was only Sylivia Pilet's Newton on the misato tracker that dropped off after 3 or 4 days. I'm now tracking all 5 NPDS servers again.

My npdscmd.txt contents (and time-stamp) look fine. NPDS servers obtained from shared trackers are not saved in this file. Only NPDS servers that actually registered with this tracker, in this case, my own NPDS server, are saved in this file.

-rwxr-xr-x 1 root wheel 70 Sep 13 06:35 npdscmd.txt

REGUP newtzilla.quadzilla.net Tahoe Newton (newtzilla.quadzilla.net)

I've activated logging, to see what kind of messages are saved. I will let the tracker run longer, a week or more with logging on. I've also looked over the .java code, where the shared server's info is gathered. I'll post some relevant chunks and comments later this week.

[morgant]

I just had to restart the UNNA.org tracker again because it had lost communication w/the shared trackers. At least this is 100% reproducible for us. I'll also take a peek at the shared server section of the java code and may add some additional logging to see what's up.

[QuadzillaNET]

Yesterday, my NPDS server newtzilla.quadzilla.net started showing as Down.

Tapping the UP button in the Tracker Client reports back "error 400 Bad Request server already exists in list" and the status of my NPDS server does not change. I'm going to say that this is normal, probably expected. Interesting to note: npdscmd.txt shows the same as above, REGUP, though my server is listed as Down in the tracker.

Today newtzilla.quadzilla.net has completely disappeared from my tracker.

Tapping -[]- disconnect in the Tracker Client removed me from the npdscmd.txt file. I waited a couple minutes and then tapped UP. This seems to have brought me back to being Up this morning and I am now tracking all 5 NPDS servers.

I will keep running my tracker as it is, but inserting additional logging of the mSharingInfoVector and mHostInfoVector vectors is probably a good idea.

"Roger Roger … What's our vector Victor?"

[splorp]

Thanks for playing around with this, guys. Perhaps I should leave my tracker up with logging as well. It’s sitting behind my router on the LAN at the moment, but it fetches the shared trackers fine. It might be interesting to seeing whether the same issue affects it.

[morgant]

On July 8th I found a try/catch in npdstracker::validateServers() that didn't throw an exception (line 1613 of npdstracker.java) and added logging of the exception (see this commit in my debug branch). According to the logs, it's been running since 2013-07-08 at 13:47 EDT and sharing successfully with misato.chuma.org (the other two it shares with have been offline for the duration) until 2013-07-14 at 05:18:51 EDT when it started getting the following exception: java.net.ConnectException: Operation timed out. This exception has continued for every connection attempt since, even though at current check misato.chuma.org is online and is showing shared servers that are registered with my tracker.

I have reset my tracker server and log file and we'll see how long it stays up this time (hopefully more than six days).

[splorp]

Well, that’s more information than we had before … even if it’s not much.

[morgant]

@pgeremia I'm hopeful that some of your changes migrating from Thread to Executor will resolve this issue.

[pgeremia]

We shall see! I am going to keep working on this to see if I can help out.

Sent from my iPhone

On Apr 27, 2015, at 12:21 PM, Morgan Aldridge notifications@github.com wrote:

@pgeremia I'm hopeful that some of your changes migrating from Thread to Executor will resolve this issue.

—
Reply to this email directly or view it on GitHub.

[morgant]

Much appreciated!