New Stats possible problem

[The_Celt]

Hi. Thanks for turning on the stats again, but I see a small problem. My active clients for past 7 and 50 days are a litle to big. My max clients should be 15, not 30 and 26. My stats page can be viewed at "http://fah-web.stanford.edu/cgi-bin/mai ... e=The_Celt".

Thanks,

[The_Celt]

Hi, and no, I have not reconfigured a lot in the last week or so. I read another recent post about this issue, and that is not the answer for me.

Thanks again.

[VijayPande]

thanks. I think we know what's going on here. From what I can tell, it's not an issue with the stats system (which is just reporting what it's told), but with the v5 server code's output on a small technical point, that's screwing up this output. We're looking into this.

[VijayPande]

PS On a second look, I think even the server code and stats is working fine in your case. Is there any chance there is someone else using the same name? If you PM me your IP address, we can check in your case. (others with this issue: please don't PM me with your IPs since we only need one case to test and more cases will just fill up my PM box and slow us down).

[ChelseaOilman]

I think I see the glitch and have reported it.

[VijayPande]

ok, upon further investigation, I think we do see an issue with the v5 server code and how it's reporting the CPUID's. We're on it.

[The_Celt]

Vijay, thanks. I haven't checked back for a few days now. If you still need my IP to check, tell me how to PM you with it.

[ChelseaOilman]

If you still need my IP to check

Pande Group knows your IP address already. On the page the mods can view, I see 14 active clients in the last 90 days. I have a feeling your just going to have to wait out the 7 day and 50 day period on the public stats page before your numbers will be reporting correctly.

[The_Celt]

OK, Thanks for the answer. I've been folding for a long time now, waiting 50 days is no big deal. Thanks for getting the stats back and working.

[Pette Broad]

Yes, I reported this a few days ago in another thread. I've no idea if this has been fixed yet, but here's my figures. The 50 day figure has dropped by one but the 7 day figure has spiraled up to 100 (It was an extremely optimistic 75 a few days ago ) In fact it's gone up 5 clients this morning. I have 56 clients running at this moment but the 7 day figure could be 59. In practical terms it makes no difference my points but it means that I take the Total Active CPU count of 430581 with a rather large pinch of salt


Date of last work unit 2010-02-11 05:03:16
Total score 17335976
Overall rank (if points are combined) 196 of 1355015
Active clients (within 50 days) 134
Active clients (within 7 days) 100

[VijayPande]

Thanks. This is a known problem, with a known cause (v5 server code minor difference in reporting CPUID's), that Joe has fixed, and we are rolling it out shortly.

Note that it will take some time for this to work itself out in the stats (since I don't want to delete old IDs, so we will wait for these to no longer be active), but I think it's soon a non-issue.

[larryb63]

Just a report that within the past week, I've seen my 'active clients' change, but my hardware has remained constant. I'm running a number of various clients (CPU, GPU, SMP) for two teams (Mozillazine and KnoppMyth) as larryb_home.

MZ Team: 6 CPU clients (3 windows/3 linux) 3 GPU clients (windows)
KM Team: 3 linux CPU clients, 1 linux SMP client (quad core)

The Stanford Stats page shows 14 active clients for MZ and 7 for KM. It looks like the KM number is correct and accounting for the quad core.

I believe that the ProtoMol cores may be causing this. I saw this 'client error' before when the ProtoMol core accidentally leaked out last year. I'm not sure, but if there's something I can do to help figure this out...let me know.

[7im]

The CPUIDs related Bug Vijay mentioned above is what caused your active clients list to double. Once for the buggy IDs, and once again for the corrected IDs. As Vijay mentioned, the numbers will normalize as the old incorrect numbers drop off in 7 days and 50 days.

Protomol WUs are coming from the v5 coded servers mentioned above. You're right on track describing the bug. It's the one they're already working on/fixed.

[larryb63]

7im: Thanks for the quick reply.

I've already seen that the KM client number has dropped, meaning that it should hopefully level out at a total of 4 clients (3 CPU, 1 SMP).

Is there any guideline as to what constitutes a 'client?' From what I see happening, it should level out as each FAH process being a client, regardless of how many cpu/gpu cores running that process. ??

From what I can figure out...you have the hex Stanford assigned 'UserID' and the user assigned 'MachineID'. Each unique setting of those two codes constitutes a client. The passkey is just for security issues, right? Where/how does the CPUID get involved in the mix? Is this CPUID the processor's internal CPUID or an assigned one from Stanford? Just curious on the internal tech details.

[7im]

CPUID is a combination of the fah UserID and MachineID, as you might have guessed.

Some of the more technical things are detailed in the FAH WIKI. For example... http://fahwiki.net/index.php?title=CPUID