Windows Client Errors

[GenomeX]

Hi Ive been testing out the client to see what it is like and Ive been having some difficulty.

When left on returning each block when completed the client runs smoothly. Though when I turn off my net connection (dial up) and return to send off as a whole test the program crashes at about 3 blocks completed.

I have had this happen many times now and as I leave my computer on 24/7 it is very annoying.

My specs are:

Windows 2000 SP2
AthlonXP

Here is the log:

[Sun Jan 12 18:31:20 2003] login successful
[Sun Jan 12 18:31:21 2003] n.high = 471168 . 33 blocks left in test
[Sun Jan 12 19:04:40 2003] logging into server
[Sun Jan 12 19:04:41 2003] login successful
[Sun Jan 12 19:04:42 2003] n.high = 523520 . 32 blocks left in test
[Sun Jan 12 19:40:03 2003] logging into server
[Sun Jan 12 19:40:04 2003] login successful
[Sun Jan 12 19:40:05 2003] n.high = 575872 . 31 blocks left in test
[Sun Jan 12 20:11:54 2003] logging into server
[Sun Jan 12 20:11:54 2003] login successful
[Sun Jan 12 20:11:55 2003] n.high = 628224 . 30 blocks left in test
[Sun Jan 12 20:45:14 2003] logging into server
[Sun Jan 12 20:45:14 2003] login successful
[Sun Jan 12 20:45:15 2003] n.high = 680576 . 29 blocks left in test
[Sun Jan 12 21:17:51 2003] logging into server
[Sun Jan 12 21:17:52 2003] login successful
[Sun Jan 12 21:17:52 2003] n.high = 732928 . 28 blocks left in test
[Sun Jan 12 21:51:51 2003] logging into server
[Sun Jan 12 21:51:52 2003] login successful
[Sun Jan 12 21:51:53 2003] n.high = 785280 . 27 blocks left in test
[Sun Jan 12 22:00:06 2003] block processing paused
[Sun Jan 12 22:00:39 2003] block processing resumed
[Sun Jan 12 22:28:27 2003] logging into server
[Sun Jan 12 22:28:27 2003] login successful
[Sun Jan 12 22:28:28 2003] n.high = 837632 . 26 blocks left in test
[Sun Jan 12 23:00:18 2003] logging into server
[Sun Jan 12 23:00:19 2003] login successful
[Sun Jan 12 23:00:19 2003] n.high = 889984 . 25 blocks left in test
[Sun Jan 12 23:40:28 2003] block processing paused
[Sun Jan 12 23:40:29 2003] block processing resumed
[Sun Jan 12 23:43:31 2003] block processing paused
[Sun Jan 12 23:43:38 2003] block processing resumed
[Sun Jan 12 23:43:44 2003] block processing paused
[Sun Jan 12 23:43:44 2003] block processing resumed
[Sun Jan 12 23:45:16 2003] block processing paused
[Sun Jan 12 23:45:19 2003] block processing resumed
[Sun Jan 12 23:45:23 2003] block processing paused
[Sun Jan 12 23:45:29 2003] cache cleared
[Sun Jan 12 23:45:38 2003] cache cleared
[Sun Jan 12 23:45:43 2003] connecting to server
[Sun Jan 12 23:45:44 2003] logging into server
[Sun Jan 12 23:45:45 2003] requesting a block
[Sun Jan 12 23:45:46 2003] got proth test from server (k=33661, n=2215224)
[Mon Jan 13 09:22:42 2003] got k and n from cache
[Mon Jan 13 09:22:42 2003] restarting proth test from cache (k=33661, n=2215224) [10.6%]
[Mon Jan 13 09:36:46 2003] logging into server
[Mon Jan 13 09:36:47 2003] login successful
[Mon Jan 13 09:36:48 2003] n.high = 254725 . 39 blocks left in test
[Mon Jan 13 10:10:39 2003] logging into server
[Mon Jan 13 10:10:39 2003] login successful
[Mon Jan 13 10:10:40 2003] n.high = 305670 . 38 blocks left in test



As you can see I have tried clearing the cache but to no avail. I thought this may work as cache errors are being returned.

Also when I restore the client after a crash it shows completed blocks as 0 :(. Am I being honoured points for these blocks or are they lost?

[dmbrubac]

Firstly, the SB client does not support being disconnected in that it won't download a bunch of tests and crunch away. The server needs to know you are still out there. It will reissue your work as early as 5 days after not hearing from you. There are other reasons too I'm sure. Read the thread about what is wanted in the new client for more info.

Secondly, SB uses your computer in some fairly uncommon ways. If you are crashing, you may have marginally bad memory.

Thirdly, the blocks done counter resets after stopping and restarting the client. It does not query the server. That being said, the 'blocks' are lost anyway, because blocks per se don't really exhist. They are just a way to divide up the real work (the full test) so that slow machines report work down before the 5 day timeout. The test isn't done and the results aren't useful until the full test is complete.

You do however still get credit for 'blocks' - click your user name in the client to see your stats.

[GenomeX]

It is not bad memory for Im sure. It is original Samsung PC2700 that is running at 150mhz and I have had it up to 180mhz during benchmarking :). This machine is not the problem the machine can prime for more than 24 hours and runs other DC projects 24/7 at 100% CPU load. :) I have new RAM in there to test at the moment though ;)

I am connected every 6 hours or so and when I set it to send blocks back as completed THERE IS NO PROBLEM AT ALL. Only when it is set to not upload during the 6 hours of being offline does it ever crash. When running online it also is quite happy to run alongside other programs no problem and then when all other programs are closed it crashes. That has got me thinking its some bug on my machine causing it.

[smh]

As you can see from your log file, the client continued where it left off when it crashed. The only time when the test restarted was when you cleared the cache yourself!

[dfamily]

Genome, when you say crash, how do you mean. Does the entire machine freeze up or it appears that the client is no longer working and do you get any kind of an "error" message in the client window? I've had a 2200xp freeze up on me for no apparent reason but I think it was a hard drive issue.

Generally speaking, when it can't connect to "upload a block" it is just saying it can't report in and keeps grinding. The log will report:

[Tue Dec 31 06:52:19 2002] n.high = 462567 . 23 blocks left in test
[Tue Dec 31 07:29:48 2002] temporarily unable to connect -- block added to submit queue
[Tue Dec 31 08:07:18 2002] temporarily unable to connect -- block added to submit queue
[Tue Dec 31 08:44:27 2002] logging into server
[Tue Dec 31 08:44:27 2002] login successful
[Tue Dec 31 08:44:27 2002] n.high = 660810 . 20 blocks left in test

As long as your block percentage is changing in the client window, you're still grinding...when you reconnect, the next time it's ready to report, it will do so.

Now, when it's finished with a test and can't connect, you'll get a different error message in the client window. It will continue to try every 20 minutes by default. When you reconnect, let it cycle itself and it will automatically try and connect. (If you stop and exit, when you restart it will pick up at it's last saved point and redo that part of the test - a real bummer on a older K6) The client window will say something about an Error 0 but that's not a prob...if it shows an Error 3, then watch it when it hits the end of the 1200 minutes. If your connected and it doesn't, then you may have to clear the cache (this happened to me twice so far but don't know why).

Hope this helps...i've got 5 machines running on a sometimes connected wire (the office fax line) and if they all start and finish about the same time, I only have to make sure they're connected for a little bit every 4 hours or so...

[GenomeX]

Originally posted by smh
As you can see from your log file, the client continued where it left off when it crashed. The only time when the test restarted was when you cleared the cache yourself!

Yes but this does not explain the crash.

dfamily it was a cache error to begin with, which was fixed via the clearing of the cache.

Now only the program is crashing leaving it at the desktop and causing a windows error in program prompt.

[Cowering]

Well, I might as well air out the problems people on my team have been having that I was not able to fix.

1. One person with WinXP can leave his machine running SB for days with no problems.. but then he'll come back from a nights sleep to a thrashing machine.. cause SB is suddenly using 800+ Meg of RAM on his 1 gig DDR RAM machine (p42.0ghz)

2. The client does not close sockets.. I know this bug is fixed and I anxiously await the new release. This bug screws with norton personal firewall and a few other lesser know net sharing clients.

3. No proxy support.. I cannot figure out how to get the windows client to use a firewall right.. I've opened port 1717 and the clients still can't get a block from the server. I've got permission to add SB to the pushed startup script of several hundred boxes if I can figure out how to get blocks to them.

4. If I insert the sb client into a 'new' box via a startup script, is there a magic way to make sure it gets a fresh block? I've tried putting sb.exe into the startup folder and letting regedit insert all the registry keys (after deleting the K and N keys) but the client never autostarts like it should.

[dmbrubac]

Cowering:

WRT #3, use sbp.pns.net for host and 22 for port. I had the same problem on my firewall until I set it up this way.

[Cowering]

I'll try that.. please tell me i did not not that in a FAQ somewhere

[dmbrubac]

I see your account age is at 33 days. I don't know if the FAQ had been removed by then or not. I got the info from the FAQ, but it disapeared shortly after I started (39 days)