Yup, I just uploaded 12 in 9 mins, its a good tweak, even if not logical.
Andy

ROM

So many people have replied that "this works," it is not nefarious... It is just that there are not enough process/threads/child processes that can be spawned... They are caught in the 300 second network timeout and hand off to the other server that also has a 300 second network timeout... The Proxy has additional threads that can be spawned... So it is not nefarious... It is configration, it is the same IP address that is only handed off...

I don't think it is as nefarious as you all seem to think it is.

Basically the proxy headers are ignored by the web server and the request is processed just like it was any other web request. It just so happens that the cgi directory is mapped to the same location on both servers. So when an HTTP request is made for the file_upload_handler it executes the file_upload_handler on galileo instead of kryten.

The upload/download directories are also mapped to the same locations as well. So things should work.

I have sent mail to Matt and Jeff letting them know what is going on.

Now whether they let it stay that way is another thing.

---

Please consider a Donation to the Seti Project.

This works because the scheduler and the upload/download server used to be on the same machine. Now they are split into two. Even though both have the same functionality, one server is doing one thing while the other is doing the other. What decides? DNS. Both can still act as either scheduler or upload/download server...

...until tonight.

Yeah I just switched it off on galileo. People trying this proxy are now getting 404 errors. Not that I'm mad or nothing - This is hardly nefarious activity as it isn't a security risk. But it's putting more traffic on the UCB LAN when it should be on Cogent. Sorry the Cogent pipe is clogged. It'll clear.

- Matt

---

-- BOINC/SETI@home network/web/science/development person
-- "Any idiot can have a good idea. What is hard is to do it." - Jeanne-Claude

Probably more worrying is that the waiting for validation queue is growing at ~4000/hour and at a time when the Cogent pipe is blocked.
Andy

ROM

So many people have replied that "this works," it is not nefarious... It is just that there are not enough process/threads/child processes that can be spawned... They are caught in the 300 second network timeout and hand off to the other server that also has a 300 second network timeout... The Proxy has additional threads that can be spawned... So it is not nefarious... It is configration, it is the same IP address that is only handed off...

Others in the thread were making statements to the effect that it was a security issue or bug in the code. It was just how things were configured.

---

----- Rom
BOINC Development Team, U.C. Berkeley
My Blog

This works because the scheduler and the upload/download server used to be on the same machine. Now they are split into two. Even though both have the same functionality, one server is doing one thing while the other is doing the other. What decides? DNS. Both can still act as either scheduler or upload/download server...

...until tonight.

Yeah I just switched it off on galileo. People trying this proxy are now getting 404 errors. Not that I'm mad or nothing - This is hardly nefarious activity as it isn't a security risk. But it's putting more traffic on the UCB LAN when it should be on Cogent. Sorry the Cogent pipe is clogged. It'll clear.

- Matt

=====================
Problem is (with very limited exceptions), the Cogent pipe looks like it did for the week we were down.

It doesn't look very blocked, or am i looking at the wrong graphs?


gigabitethernet0_0: 66.250.4.74: ucb-cogent p2p



gigabitethernet0_1: 66.28.250.121: cogent ssl net

---

Grant
Darwin NT

Normally after an outage i can't even contact the Schedulers, let alone upload or download work.
At present i've got no problems contacting the Schedulers, i'm just not able to return results (or close enough to not being aable to- only 5% of my results ready to go have been returned in the last 12 hours) & taking lots of retries to download new Work Units to process.

---

Grant
Darwin NT

It doesn't look very blocked, or am i looking at the wrong graphs?....

I think you're looking at the right graphs, but if my memory is correct, when Boinc/Seti ans Seti Classic both ran normally, wasn't the average closer to 50 M. I believe Classic has been temporarily disconnected to give us more room in the pipe, but I don't believe it is happening if my memory is correct.

I know that you can never get the full amount, gigabit, down the pipe but isn't 30 M ave, 85 M peak a little bit low?

I'm electronics engineer so know nothing about networks, or so I keep getting told.

Andy

---

It doesn't look very blocked, or am i looking at the wrong graphs?

It looks like something is wrong to me but I don't know much about these things.

See how it is at "30 M" when we were shut down. Now it is up and running and it is back down to "30 M".

---

This may show it better.

---

I too tried the proxy server ip address and it does not work for me here in San Diego, CA.

Jammy

---

Team SETI.USA

I too tried the proxy server ip address and it does not work for me here in San Diego, CA.

If you check half a dozen or so posts up you'll see where Matt says he's disabled that option.

---

Grant
Darwin NT

I too tried the proxy server ip address and it does not work for me here in San Diego, CA.

If you check half a dozen or so posts up you'll see where Matt says he's disabled that option.

Oooo . . .sorry, my faux pas!

---

Team SETI.USA

I too tried the proxy server ip address and it does not work for me here in San Diego, CA.

If you check half a dozen or so posts up you'll see where Matt says he's disabled that option.

Even before it was disabled it still didn't work here. I got 2 thru on upload, but they won't report.

Yoinked

---

According to the graph, Cogent is clearly not clogged. It obviously can get up to 80 Mbps.

See the deep dip in the graph at 1600 - that's probably when Classic got shut off. Then they opened the Seti-Boinc floodgates right after 1600, and use went up to almost 50. Then the automount problem kicked in, and use dropped at 1800. They got that fixed around 21:30 and use shot up to 85. Then something filled up again, and use came down to 40-45 range and stayed there. I bet Classic was turned back on about midnight, so the level stays around 35-40, even while Boinc is down for Wednesday's backup. When the backup is finished at 11:30, use goes back up to 70 until that elusive something fills up again, and use drops to lowest point and stays there. It appears that Boinc is using only a few mbps on top of Classic because something is filled up and is only letting a dribble through.

That's my armchair analysis of what's happening. Good luck, Matt, finding the bottleneck. By the look of the graph, it sure isn't a clogged Cogent line. When I used the proxy, U/L and D/L went swell. There was no delay when we got through on the camput net, and the WUs were processed right away, suggesting no problems getting WUs through the processes. Could there possibly be another slow switch/router or NIC? Keep smilin' - we are all rooting for you.

Whit

According to the graph, Cogent is clearly not clogged. It obviously can get up to 80 Mbps.

See the deep dip in the graph at 1600 - that's probably when Classic got shut off. Then they opened the Seti-Boinc floodgates right after 1600, and use went up to almost 50. Then the automount problem kicked in, and use dropped at 1800. They got that fixed around 21:30 and use shot up to 85. Then something filled up again, and use came down to 40-45 range and stayed there. I bet Classic was turned back on about midnight, so the level stays around 35-40, even while Boinc is down for Wednesday's backup. When the backup is finished at 11:30, use goes back up to 70 until that elusive something fills up again, and use drops to lowest point and stays there. It appears that Boinc is using only a few mbps on top of Classic because something is filled up and is only letting a dribble through.

That's my armchair analysis of what's happening. Good luck, Matt, finding the bottleneck. By the look of the graph, it sure isn't a clogged Cogent line. When I used the proxy, U/L and D/L went swell. There was no delay when we got through on the camput net, and the WUs were processed right away, suggesting no problems getting WUs through the processes. Could there possibly be another slow switch/router or NIC? Keep smilin' - we are all rooting for you.

Whit

I like your armchair analysis. Wish Matt would look at the graph and come to the same conclusion that the cogent link is not all blocked up. Clearly it ain't. Clearly something in SETI/BOINC land is fubar!

According to the graph, Cogent is clearly not clogged. It obviously can get up to 80 Mbps.

See the deep dip in the graph at 1600 - that's probably when Classic got shut off. Then they opened the Seti-Boinc floodgates right after 1600, and use went up to almost 50. Then the automount problem kicked in, and use dropped at 1800. They got that fixed around 21:30 and use shot up to 85. Then something filled up again, and use came down to 40-45 range and stayed there. I bet Classic was turned back on about midnight, so the level stays around 35-40, even while Boinc is down for Wednesday's backup. When the backup is finished at 11:30, use goes back up to 70 until that elusive something fills up again, and use drops to lowest point and stays there. It appears that Boinc is using only a few mbps on top of Classic because something is filled up and is only letting a dribble through.

That's my armchair analysis of what's happening. Good luck, Matt, finding the bottleneck. By the look of the graph, it sure isn't a clogged Cogent line. When I used the proxy, U/L and D/L went swell. There was no delay when we got through on the camput net, and the WUs were processed right away, suggesting no problems getting WUs through the processes. Could there possibly be another slow switch/router or NIC? Keep smilin' - we are all rooting for you.

Whit

I like your armchair analysis. Wish Matt would look at the graph and come to the same conclusion that the cogent link is not all blocked up. Clearly it ain't. Clearly something in SETI/BOINC land is fubar!

Like the validator queue again? It's on the rise!

---

I noticed that my home machines sometimes start the upload, the whole data file is uploaded but then stalls, times out and goes in to retry mode. Damn almost there.