Welcome MerePeer

[Fozzie]

Good to have you on SoB with us. :cheers:

[LAURENU2]

I will 2Nd that:thumbs:

[Bok]

Welcome!

Great to have you crunching with us. We are slowly getting close to that magical 1000M for our rate!!

Bok

[MerePeer]

Thanks for the welcome mates! Here to help a bit.

Wondering: why are there significant discrepencies between the official and our stat pages for the team for some members who are static?
http://www.seventeenorbust.com/stats/teams/team.mhtml?teamID=60
http://stats.free-dc.org/new/teamstats.php?proj=sob&team=FreeDC

Examples:

15 jandersonlee 59.62 T cEMs (vs 164T on freedc stat pg)

14 rsbriggs 78.24 T cEMs (vs 80T)

27 rjkc4 22.51 T cEMs (vs 256T)

36 em99010pepe 6.57 T cEMs (vs 7T)

41 NEBOJSA 4.06 T cEMs (vs 9T)

=============================
Also, unrelated: -- is there any software to monitor nodes/make sure they havent died? (I had one die last night).

[Fozzie]

there is a switch to restart stalled clients.

"sobsvc -m"

[Bok]

Thanks for the welcome mates! Here to help a bit.

Wondering: why are there significant discrepencies between the official and our stat pages for the team for some members who are static?
http://www.seventeenorbust.com/stats/teams/team.mhtml?teamID=60
http://stats.free-dc.org/new/teamstats.php?proj=sob&team=FreeDC

Examples:

15 jandersonlee 59.62 T cEMs (vs 164T on freedc stat pg)

14 rsbriggs 78.24 T cEMs (vs 80T)

27 rjkc4 22.51 T cEMs (vs 256T)

36 em99010pepe 6.57 T cEMs (vs 7T)

41 NEBOJSA 4.06 T cEMs (vs 9T)

=============================
Also, unrelated: -- is there any software to monitor nodes/make sure they havent died? (I had one die last night).

Hi MerePeer,

to answer your first question...

it's because when you move teams the points don't move, BUT there is no way to tell this from the stats feed produced by SoB. I've long asked for it but to no avail...

so if you look at mine, there is a discrepancy of around 43T which you will find if you look at the retired users at the bottom of the US-Distributed team page...

And for the 2nd question - no, not really, think you could do something???? :Pokes:

Bok :thumbs:

[MerePeer]

In another thread, PCZ said points will get removed if Foz doesnt finish a WU. First off, is a WU the countdown from 6K to 0?

[Sun Sep 3 17:39:23 2006] n.high = 8985 . 6558 blocks left in test

..

[Thu Sep 7 15:41:19 2006] n.high = 3798858 . 4449 blocks left in test


And so on this sempron 2400 its doing about 500 blocks a day which would make about a 2 week WU?

Now is it the case that if I stop that PC and have it do something else before it gets down to 0, then I will lose whatever credit I had built up for that WU when someone else picks it up (where it left off?) and finishes it?

Here's a different problem: I think when I restarted this job that my default directory was not the same. Now I have both
/z1793644
and
/var/opt/sob/z1535271

[Thu Sep 7 15:41:19 2006] n.high = 3798858 . 4449 blocks left in test
[Thu Sep 7 15:42:45 2006] iteration: 3800000/11793659 (32.22%) k = 21181 n = 11793644
[Thu Sep 7 15:43:36 2006] resolving hostname
[Thu Sep 7 15:43:36 2006] opening connection
[Thu Sep 7 15:43:36 2006] logging into server
[Thu Sep 7 15:43:36 2006] login successful
[Thu Sep 7 15:43:36 2006] n.high = 3800655 . 4448 blocks left in test
[Thu Sep 7 15:45:18 2006] internal computation error [mismatched sums]! check your memory/processor. test
will restart in 5 minutes.
[Thu Sep 7 17:55:38 2006] client process [v2.5.0] invoked
[Thu Sep 7 17:55:38 2006] priority set to idle
[Thu Sep 7 17:55:38 2006] connecting to server
[Thu Sep 7 17:55:38 2006] logging into server
[Thu Sep 7 17:55:38 2006] requesting a block
[Thu Sep 7 17:55:38 2006] got proth test from server (k=24737, n=11535271)
[Thu Sep 7 17:55:38 2006] server packet cached to disk
[Thu Sep 7 17:55:38 2006] AMD Sempron(tm detected. Enabling cpu specific optimizations.
[Thu Sep 7 17:58:02 2006] resolving hostname
[Thu Sep 7 17:58:02 2006] opening connection
[Thu Sep 7 17:58:02 2006] logging into server
[Thu Sep 7 17:58:02 2006] login successful
[Thu Sep 7 17:58:02 2006] n.high = 1878 . 6142 blocks left in test

Also -- is there a way to tell it not to grab another WU when it is done?

Thx.

[MerePeer]

Ok my (linux) startup script for sob did not set the default dir so it used /. But I have a directory there called /cache so the sb program couldnt create that file (but never complained). So every time the system boots it grabs a new block because it doesnt know it was working on one previously (because no cache file). Am I correct that there is no way to "restart" these?

/var/opt/sob# ls -ltr /z*
-rw-r--r-- 1 root root 1474316 2006-09-05 09:54 /z1794349
-rw-r--r-- 1 root root 1477992 2006-09-06 10:01 /z1823737
-rw-r--r-- 1 root root 1439932 2006-09-06 12:42 /z1519258
-rw-r--r-- 1 root root 1440432 2006-09-07 09:59 /z1523260
-rw-r--r-- 1 root root 1441940 2006-09-08 21:31 /z1535336
-rw-r--r-- 1 root root 24 2006-09-08 21:48 /z1548030

[PY 222]

Welcome MerePeer.

To answer your question, yes if you do not finish a test, then all the blocks that you have processed will go to waste.

My startup script is fairly simple and it has a crontab that will run /sob/restart-sob script in which the sb client will get started every X hours. I do not get a new z value and it happily restarts the sb client and the same test.

[Fozzie]

If you completely stop the WU and don't do anything on SoB with that cahced wu for a whole month then it will be returned to the dropped queue for reprocessing.

All you need do to hold onot that job is to process at least one block per month and then when it finally does finish you keep all your credited work.

So you don't need to concentrate on SoB on a slower box just run it every so often to keep the wu alive.

[PCZ]

If you have send intermediate blocks on then you will be getting credit for a WU before it is finished.
If you don't finish that WU then the credit gets removed.
Not straight away but some time in the future when another person finishes it.

If you loose a WU after having only doing a few blocks then it doesn't matter much but if you have done a lot of blocks then that will make a big dent in your credit when the points get taken away.

WU's can be salvaged , if you have the z files.
It is a known problem with SOB and the workaround is documented.
http://www.free-dc.org/forum/showthread.php?t=7434
Not sure if it works in linux though.

I suggest you log into the SOB webpage go to preferences then click here in the pending test management section.

You will see a list of all assigned to you.
Try and salvage the ones with a high percentage done.
Tests with only a few percent can be expired, if you are sure they won't be finished by you.

[MerePeer]

Thx for the replies. The documented recovery solution isnt applicable to linux, although I did determine that the first 4 bytes of the linux {cache} file are the N and the second 4 are the K, the last 4 bytes I couldnt pin down. So since I couldnt generate a cache file for those z files I went to my test mgmt page and expired all of them; just one system had the issue anyway. A lesson learned. :looney:

[Fozzie]

I'll check that Monday to see if I can recover that lost wu

[Fozzie]

Used the process from the link provided and viola instead of starting from 0.0% and losing all those points I now have the 7.8% back and continuing. :cheers:

Just picked up one that the dual lost when switching from running the client twice to running as a service.

21% completed WU saved :cool: