Deltavee:
Sorry to see you go. You have been a major contributor. The losers are the researchers.
Cheers

I've done a little bit of network tracing on the connection problems I was having a couple of weeks back, and I could see that the TCP connections were made, but I was getting TCP resets back from the WCG server end.

There is some data coming over these connections but without fiddling with TLS capture and decryption I can't see whats in the response.

If this is a 503 error, it would indicate the stack is working up to the httpd application layer for us to get that error.

If I was working on this, I'd be looking at the capacity of the httpd processes being used as there are usually settings that affect max processes\threads\sockets etc.

I've never worked on anything of this scale however, only a few thousand users or so which is tiny compared to the scale of WCG.

next problem:
just for testing purposes, I recently added another PC, and eventuelly I got some tasks. I then wanted to assign a profile to this PC, but looking up the device manager showed me that this new PC is not listet.
Which shows that this does not work either

My frustration level about WCG has reached a maximum right now. How come it took them months since they left IBM, and at least half of the whole thing is not working properly.

And the next frustration is: no word whatsoever from the project team I am afraid that it is not clear to them that we all spend our time as well as part of our constantly increasing electricity bill for this project. So they could at least tell us from time to time what's going on and how much longer it weill take them to get all the current problems straightened out

Well, sharnet.ca is the universities' service provider in the eastern provinces of Canada where Krembil is part of (they are in Toronto). And there certainly is some form of load balancer/aggregator in front of the actual servers, so that's what you're likely seeing in your captures.
Me neither, but I am doing IP networking for about 35 years now. And for the last two months I have been saying that this is not a "bandwidth" issue, but something more server side. The httpd server could be one of the possibilities, but an also often overlooked issue that I have seen (for example when small Internet startups here in Silicon Beach are trying to "scale up" just before or after they sold out) is the number of concurrent file handles of the file system or a similar issue on the database server (cluster) and access to the database.

Yes, the response on a failed connection is a 503 error and that corresponds rather perfectly with those 107 bytes (no more, no less ) that you see as "transferred" when the connection actually doesn't get established. This seems to indicate that the IP stack in fact is making a connection, but no data is forthcoming from either the database or the filesystem.

But this is an issue that one can only properly diagnose on the "inside" of the setup, as that is communication that you can't capture from the outside. So let's hope that the powers to be at WVG's tech team can switch their radars from long range mode to close-in combat mode and finally come up with a lasting fix for this. It's two days until the 1 year anniversary...

Ralf

Yes, validated, sent to researchers and counted.
External sites are extracting stats but they do not show here.
See also other recent threads:
https://www.worldcommunitygrid.org/forums/wcg/listrecentthreads

Paul.