FAO Howard

[TyOI]

TyOI

The power outage yesterday has only made a bad problem even worse

any chance you can comment or advise on these problems

when this new client was released it took almost 3 hours before one could connect to obtain a download, then it took three attempts before a working download could be obtained - now one cannot connect to the server to upload - this client is so fast those with fast connections seem to be hogging the servers so us with poorer connections cannot even get a connection - this is obvious by the amount of work that is going up

these problems of being able to download a new client and upload the results need to be looked into and the above is the reason i do not commit the full power of the S-MDC to the project as i would never be able to upload its work with the type of connection i have

Regards
tom

[Aegion]

Originally posted by TyOI
TyOI

The power outage yesterday has only made a bad problem even worse

any chance you can comment or advise on these problems

when this new client was released it took almost 3 hours before one could connect to obtain a download, then it took three attempts before a working download could be obtained - now one cannot connect to the server to upload - this client is so fast those with fast connections seem to be hogging the servers so us with poorer connections cannot even get a connection - this is obvious by the amount of work that is going up

these problems of being able to download a new client and upload the results need to be looked into and the above is the reason i do not commit the full power of the S-MDC to the project as i would never be able to upload its work with the type of connection i have

Regards
tom

Read some of the other threads. Basicly he expected to have sites mirroring the download when the new client was released, but it didn't happen. He's trying to get some other mirrors from volunteer with sufficient bandwith for future client and protein switchovers. The current problem should be gone in a day or so. What's happening is everyone is attempting to upload their structures at once and the server is getting overloaded. A power outage of this scale is highly unusual and I would not expect that the server would every be down for a significant period due to the same reason again. In another day or so, the server should be functioning normally again.

[jlandgr]

Originally posted by TyOI
now one cannot connect to the server to upload

In fact, I just uploaded 100 generations just now without a delay, so if you're reading this, try now (it's always best, in backlog situations like these, to wait till the middle of the night in North America to upload data ).
Jérôme

[the-mk]

I was able to upload 2/3 of my offlinework without a delay.
The other 1/3 were destroyed by that silly client about 200 generations are lost (previos generation missing )

At least my was about 200,000 points, so that's not bad!

[jlandgr]

Originally posted by the-mk
about 200 generations are lost (previos generation missing )

Do you still have a copy of the directory?
If yes, try rescuing some generations with the "-purgeuploadlist" command (see readme1st.txt). Paratima posted about this, see http://www.free-dc.org/forum/showthr...&threadid=3914

Jérôme

[tpdooley]

from July '02 to Jan '03 I used 56k dialup on all my 16? folding machines. (numbers vary as some of those that claimed they'd fold decided otherwise later on..
most of the machines have been moved to cable modem access since then and things have gone reasonably smoothly except for the last two changes.. (may Howard be whipped with a wet noodle repeatedly for switching compilers and not testing it.. and this time because of a power outage - which falls under the "totally unforseen circumstance" category..

so while the high speed connection folks have got the upper hand on uploading data with this client.. we're still dealing with the same problem you're describing.. i.e. only some of the stored data being uploaded before a timeout..

(someone offer to exchange those measly 4xxMhz cpus for higher speed versions! announce the cost of each.. and I'm sure a number of us will donate to get them upgraded to handle the increased workload..

[the-mk]

Originally posted by jlandgr
Do you still have a copy of the directory?
If yes, try rescuing some generations with the "-purgeuploadlist" command (see readme1st.txt). Paratima posted about this, see http://www.free-dc.org/forum/showthr...&threadid=3914

Jérôme

I've *.zip-ed this directory...

I tried that "-purgeuploadlist"-thing but it didn't work (I "purged" it until the last fileset). As I tried to upload them after the server was back up the client told me he wants to upload fileset 1/96, so where are the other 100 generations? That machine was running about 20 hours offline, produced about 200 generations (dfGUI said), but only about 100 were buffered, but I couldn't upload none of them

If anybody wants the zipped copy of that directory PM me

[bwkaz]

Originally posted by TyOI
this client is so fast those with fast connections seem to be hogging the servers so us with poorer connections cannot even get a connection

What?

That doesn't even make any sense. I'm assuming there are a preset number of concurrent connections that either the database or the web server can handle at a time. Assume you're trying to get one of them, and someone else currently has one.

Now, would you rather have that other person be on dialup, where it'll take a great deal longer to actually perform the upload, or on some sort of broadband, where it'll take much, much less time?

Seeing as that was a rhetorical question, I'll answer it for you (). You want them to be on broadband. This used to be the problem with FilePlanet, actually (and it still might be, but I don't wait around for Counterstrike updates anymore, so I haven't noticed). Their servers have a connection limit of (say) 200 users. When the server first comes up, those 200 slots get filled with a mixture of dialup and broadband users. However, as the broadband users finish their download (which happens in a much shorter time than it would take if they were on dialup), other users take their place. These other users are also a mixture of dialup and broadband.

Eventually, you end up with a server where 95% of the connections are from people with 56K modems, and the other 5% are from people that can actually get in, get the file, and get back out. And the lines waiting for a download to even begin start to get much, much longer.

Assuming the ratio of dialup to broadband users is 1:1 (it's not, but oh well), and the ratio of dialup bandwidth to broadband bandwidth is 1:20 (on average), then if you start with 200 connections, half of them will be dialup users. Those slots will be taken up for 20 time slots (how long a "slot" is depends on the size of the download). The other half will be taken up for 1 time slot. After that time slot is done, that half will be available, and a 1:1 ratio of dialup to broadband users will fill it. The broadband users (by this time down to using only a quarter of the total slots) will again finish in one time slot, and those spots will be filled with a 1:1 ratio again. This continues.

The original half of the users won't give up their slots until the broadband users are down to one (2^20)th of the number of slots they had before. Basically, this is zero.

Of course, it's good for the dialup users, because they have more slots opening up faster (until near the end, anyway).

Your problem isn't the broadband users. They're doing their uploading and getting out of the way fairly quickly (if they can ever connect, that is). It's pure and simple the number of people all trying to upload.

Which is why I'm waiting until late tonight to try.

[Brian the Fist]

As I mentioned elsewhere the bottleneck in this case was the database servers actual CPU horsepower, the connections were not maxxed out most of the time (though there were SOME of the time - like when you get Unable to Initialize databases upon trying to log in on High Flyers).

Anyhow, since last night it seems to be down to normal again so there should be no more delays.

As for missing previous generation messages, remember you must interpret it literally - it means that the last file the server received from your 'simulation' does not match with the generation you just uploaded (+1). It could be because you already uploaded it successfully but the client received no acknowledgement (perhaps because the server was bogged down and took to long to send a reply - the clinet gave up waiting and carried on with its business), and so kept the file, hoping to upload it again successfully. In this case, just ignore the message - the file will be wiped and it will proceed to upload the next generation. Alternatively it could mean that say, the last generation you uploaded was 8, and now you are trying to upload #10. Thus gen 9 is missing. In this case all remaining work in that 'simulation' is useless as a step is missing, and it must all be discarded.

The good news is that in most cases, I think, the message should be caused by the former reason. The only way the latter could happen is if the client uploaded gen 9 and thought it was successful, when in fact it was not. This should never happen and if it does, there is a bug somewhere in the client/server communication. Again, if anyone suspects this may be a bug, please provide as much info as possible. But first try to rule out the former possibility...

[TyOI]

TyOI

guess the servers must still be busy as its taking the earth to upload work - most times it starts to upload and then after a cpl of units have been uploaded decides its had enough then i get the cannot connect to server message and the client goes merrily on its way folding - annoying this as i have been babysitting this uploading for the past 12 hours and still have not managed to send half of the work

regards
Tom

[IronBits]

I haven't had any problems uploading instantly...
I heard some areas are still having network problems.
tracert to his server and see if you can spot some high lags along the way.