Uploading a completed WU, after net disconnect

[Aaron_Stalling]

Hello All,
What is the preferred method for uploading a completed WU in the following scenario:
I tether my phone to my computer to fold, but I disconnect the phone tethering in the afternoon . Overnight I let my computer continue to fold; in the morning both the CPU and GPU have completed the work units, but have not sent the WU yet.
Typically the WU says it will try a resend in X amount of hours, should I:
Leave the F@H client alone until it automatically sends the completed WU and then begins a new WU.
Close the F@H client and the start it again, hoping it resends the WU unit when it restarts.

[Joe_H]

Restarting the client should cause it to send the WU's in. This is assuming you are folding on Windows. With Linux and OS X the folding client is running in the background as a service by default. It takes a bit more to restart the client and reset the retry delay period.

[Aaron_Stalling]

Yep Windows 8,

Thanks.

[bruce]

If the client has finished both slots my personal recommendation would be to restart the client, though technically it's not required. Starting from the time each WU finished, the client will automatically attempt to upload the results and the time between attempts will grow progressively longer each time it tries, so the lost time before it tries again will depend on how many times it has tried and how long it has been since the last attempt. After restarting the client, it will immediately attempt to upload everything that's finished.

The client is also designed to download one WU per slot, so in addition to uploading, it will need to download a new assignment for each slot. Those downloads will also begin immediately after a client restart. Processing all four transactions at the same time (two down and two up) will saturate your modem for a while, but that shouldn't be a problem as long as you expect it.

Tethering the phone and just waiting is also an option, although as I said, some extra time may pass.

[PantherX]

If your connection can't handle 2 uploads and 2 downloads simultaneously, you may want to configure the slots so that they will start in a paused state which ensures that every time you reboot your system, the completed WUs will be uploaded but no new WUs will be downloaded until you manually enable folding on each slot.

To add the command, please do the following:
Open Advanced Control (AKA FAHControl) -> Configure -> Slots Tab -> GPU -> Under the Extra slot Options, enter the following:
Name: pause-on-start
Value: true

[billford]

Starting from the time each WU finished, the client will automatically attempt to upload the results and the time between attempts will grow progressively longer each time it tries, so the lost time before it tries again will depend on how many times it has tried and how long it has been since the last attempt.

I noticed this a while ago when my internet connection went down just after the client had downloaded a new WU but before it had finished the current one (good timing!). It was only down for around 20 minutes so not a serious problem.

I wondered at the time but didn't bother to ask- does the time between attempts increase without limit, and is there any easy way (eg adding a slot option) to limit the maximum to, say, an hour?

If not, could I suggest it for the next version? It could be useful if an internet connection failed during an extended period of unattended operation such as a long weekend or holiday away (I never turn my main machine off unless I have to).

[7im]

No, the time does not increase without limit. At max, it will still attempt to upload several times a day.

[bruce]

I don't think it increases without limit but I don't remember details about the actual range.

The fundamental reason why it increases is to protect the server from being overloaded. Suppose a server is off-line. Over some time period, more and more of the WUs that belong to that server will be completed. If every one of those clients attempted to upload immediately after the server comes back on-line, it would saturate the server's ability to handle uploads and almost nobody would successfully upload their payload. If the server is down for a long time, more and more clients would be trying to upload. By spacing out the upload attempts over longer and longer periods of time, the chances of getting through are improved for everybody. If the connection is only down briefly, the retry time can be much shorter since no large backlog has had time to accumulate.

[billford]

I hadn't really considered a server being down (or unavailable/unreachable) for any serious period of time... but it could happen I suppose.

That all makes sense, thanks.

[bruce]

Nowadays, most servers have collection servers which can accept uploads when the primary work server is unavailable. That has mostly solved the problem of a server having too many upload requests, but it still could happen under unusual circumstances so leaving that (old) functionality in the client is still a good idea.