Hi, everyone:

Does anyone know when the new upload servers will be installed and ready to except upload from finished WUs. I’m asking because I presently have 2 HadAM3Ps suspended with less than five hours to completion each. I also have a HadCM3 that is presently crunching its way through 2079. At this rate it will be ready to upload sometime Wed. night. I also have another HadAM3P that’s approaching 50% and should be done about Friday.

Fortunately, I have no problem getting replacement WUs, but, with all those suspended WUs the backups are getting to be huge. I have been considering transferring the suspended WUs to a flash drive so they won’t take up so much space in the memory. I can restore them later (when the new upload servers are online)and run them to the end.

---

No info yet.
The new power supply should be in place within 12 hours, so that will allow zip uploads. It's the trickles that may be a problem.
There are so many servers that I can't remember what goes where. :(

I've got 7 suspended; 2 for comlab (no space), 1 for atm (power supply), & 4 for oerc (no space).
We seem to be a victim of our own success. :)

2-3 hours, and Milo will be back at Oxford, so in the 12 hours following that we may hear something.
Expect to see it on the News threads, both here and on the other board.

Currently, I'm still waiting for notification about the power supply. I'll check on it shortly, but I expect that we won't see it until tomorrow. *

An order for a 20TB data server has gone in today and orders from this company normally take a week or so to be fulfilled. When it arrives it will go in a server room to which I don't have direct access and so I am not sure yet how long it will take to set up. Hopefully I will know by the time it arrives. I have some other plans in progress but this new server is the most likely to be useful in the short term.

* Edit: Looks like I was right.

Two of my models were so close to the end that they finished and are queued for upload. I'm running other projects so I don't want to suspend network activity.

How do client upload requests in this current state disrupt or impede operations?

Thanks.

The servers are down, so it won't matter. It's the 14 day time limit that will get people. If it lasts that long.

---

Backups: Here

A few weeks ago, the CPDN server was down for several days so I loaded a pile of Einstein models onto my Quad 9550. (What a delight - handfuls of models completing every 24hrs. No wonder folk coming to CPDN for the first time question the run-time of CPDN models!)

But wouldn't you know it, the Einstein server then took a dive also. As you know, Einstein rigidly enforces completion dates - 7 days, from memory - but mindful of this, they were able to extend the completion dates to allow crunchers to upload completed models after repairs to the server.

Would this simply not be possible with CPDN?

John.

... the Einstein server then took a dive also. As you know, Einstein rigidly enforces completion dates - 7 days, from memory - but mindful of this, they were able to extend the completion dates to allow crunchers to upload completed models after repairs to the server.

Would this simply not be possible with CPDN?

There are two types of deadlines that apply to CPDN:

1. The 'report deadline' that appears in BOINC Manager. This is ignored by CPDN, but has to be set to some value for any BOINC task (something to do with work shares, I assume). So BOINC Manager and the BOINC server software issue alarming messages if the deadline is exceeded but they can be ignored for CPDN. I assume that Einstein extended that task deadline (which they use).

2. When BOINC Manager attempts to upload a results file, it checks to see whether a timeout period has expired. That period is set to 60 x 60 x 24 x 7 x 2 seconds (PERS_GIVEUP) - i.e. 14 days. I've not found that setting exposed anywhere as a preference, so it's effectively hard-wired. (You could download the BOINC Manager software and rebuild it with a larger value.) It's this deadline that's the issue here. As I understand it, the behaviour is in the BOINC Manager software on our machines and there is nothing CPDN can do to change that.

[Edit: Not strictly true according to http://boinc.berkeley.edu/wiki/Client_configuration. Scroll down to command-line option --file_xfer_giveup_period. That gives the impression that BOINC Manager can be started with a new value. Is this right?]

Thanks, Iain. It's clear now - fingers crossed for a speedy return of the server, then.

John.