Clomate prediction then refused to request work because it was backed off?????

Of course!
Each time that your computer contacts the server, it gets a 1 hour and 2 minutes backoff.

So, first you set up your changes, then you wait an hour, and THEN you can make use of those changes.

Clomate prediction then refused to request work because it was backed off?????

Of course!
Each time that your computer contacts the server, it gets a 1 hour and 2 minutes backoff.

So, first you set up your changes, then you wait an hour, and THEN you can make use of those changes.

Yes but such a back off shows in the manager projects window and no such back off was showing.

As an aside, the manager appears to be playing up. I am unable to change the logging options through the manager options page and, having turned the work logging option off by editing cc_config.xml it is still active!?!

Just in case you haven't, after changing cc_config.xml, you should either restart boinc, or choose "Read config files" in the options menu to have the change take effect.

Just in case you haven't, after changing cc_config.xml, you should either restart boinc, or choose "Read config files" in the options menu to have the change take effect.

Dumbklutz, thank you - I’m an idiot.

Just in case you haven't, after changing cc_config.xml, you should either restart boinc, or choose "Read config files" in the options menu to have the change take effect.

Dumbklutz, thank you - I’m an idiot.

It does seem odd though that you can't change them within the manager. Maybe you should consider uninstalling then reinstalling the manager from scratch?

I just saw in the new work announcement threads that you released work for Macs. Didn't know that you provided a macOS supported app version as well, and that this project is solely based on Linux and Windows. Is work for Macs rather sporadic or continuous?

I just saw in the new work announcement threads that you released work for Macs. Didn't know that you provided a macOS supported app version as well, and that this project is solely based on Linux and Windows. Is work for Macs rather sporadic or continuous?

Work for all platforms is sporadic. Unfortunately at present anyway, work for MacOS is the least common of the three platforms. It is possible that Mac's decision to move to ARM CPU'S may mean new Mac's will never get any work but that is pure speculation. The batch was less than 100 tasks but as they are spin up tasks as I mentioned in the announcements thread that could well mean more tasks in the future. These tasks will also run on Linux providing the machines have the required 32bit library files.

Shouud I just be patient, or do I need to fix something?

My old machine crashed without a trace, so I bought a new one (runs Red Hat Enterprise Linux 8.2) but gets no work units.

Mon 16 Nov 2020 03:30:21 AM EST | climateprediction.net | Sending scheduler request: To fetch work.
Mon 16 Nov 2020 03:30:21 AM EST | climateprediction.net | Requesting new tasks for CPU
Mon 16 Nov 2020 03:30:26 AM EST | climateprediction.net | Scheduler request completed: got 0 new tasks
Mon 16 Nov 2020 03:30:26 AM EST | climateprediction.net | No tasks sent
Mon 16 Nov 2020 03:30:26 AM EST | climateprediction.net | Project requested delay of 3636 seconds
Mon 16 Nov 2020 07:11:39 AM EST | | Preferences:
Mon 16 Nov 2020 07:11:39 AM EST | | max memory usage when active: 25350.42 MB
Mon 16 Nov 2020 07:11:39 AM EST | | max memory usage when idle: 28519.22 MB
Mon 16 Nov 2020 07:11:39 AM EST | | max disk usage: 80.00 GB
Mon 16 Nov 2020 07:11:39 AM EST | | don't use GPU while active
Mon 16 Nov 2020 07:11:39 AM EST | | max download rate: 9600000 bytes/sec
Mon 16 Nov 2020 07:11:39 AM EST | | max upload rate: 9600000 bytes/sec
Mon 16 Nov 2020 07:11:39 AM EST | | (to change preferences, visit a project web site or select Preferences in the Manager)

I normally know I must wait a long time to get work units, but my machine has this in its ClimatePrediction description:

Maximum daily WU quota per CPU 0/day

Is this the proper value to get work units? Or must I fix something? 7.16.6

---

Is this the proper value to get work units? Or must I fix something? 7.16.6

There was a time when this was used to stop machines without the 32bit libs getting work by setting it to -1
0/day is what I see on my new Ryzen that gets work. There have been a couple of other people posting about not getting work on new Linux boxes including Les. I don't have a clue why. I know you have been around long enough that there is no point in asking most of the obvious questions. It may be worth trying increasing the number of days work requested above the default. I know Les didn't find anything in work fetch debug that was helpful.

This may not have anything to do with it, but the floating point and integer benchmarks for that powerful workstation are both exactly 1 billion ops/sec. For a PC like that, it should be > 5 billion for floating point and > 15 billion for integer. I would run a manual benchmark just to update it.

I have installed CentOS 8 in a VM and will try to see what I get when requesting work. It will take awhile since one has to jump through hoops for RH/CentOS to get boinc and boincmgr to work properly, that one doesn't have to jump through for most other Linux distributions.

I would love to run Linux in VM. Which flavour of Linux would be easy for someone to understand who has no experience of it? Easy to install and with a Windows type GUI.
I have twenty-four threads all dedicated to CPDN, sitting idle so might as well have a go at the Linux WU's.

Llinux Mint is like that.
It's based on Ubuntu.

However ...
Don't forget the need for L3 cache. The N216 model type are hungry for that, and slow WAAAAAY down it you cram too many tasks into the work load at once.
4 Megs per model.

Here is my processor according to climateprediction:

CPU type GenuineIntel
Intel(R) Xeon(R) W-2245 CPU @ 3.90GHz [Family 6 Model 85 Stepping 7]
Number of processors 16
Coprocessors ---
Virtualization None
Operating System Linux Red Hat Enterprise Linux
Red Hat Enterprise Linux 8.2 (Ootpa) [4.18.0-193.28.1.el8_2.x86_64|libc 2.28 (GNU libc)]
BOINC version 7.16.6
Memory 30.95 GB <---<<<
Cache 16896 KB <---<<<

So I should be able to run four models at a time, if I ever get any. Right?

---

Like you I am a LINUX novice and am running a VMware VM with Ubuntu 18.04.LTS on an i5 CPU (6meg L3). Seems to run OK but sometimes freezes (don't know why). In terms of throughput the optimum that I have found is 3 workunits at one time (i.e running 3cores). Single core only seems to take half the time per workunit so not quite so efficient.

This may not have anything to do with it, but the floating point and integer benchmarks for that powerful workstation are both exactly 1 billion ops/sec. For a PC like that, it should be > 5 billion for floating point and > 15 billion for integer. I would run a manual benchmark just to update it.

You are right: Here is part of my log:

Mon 16 Nov 2020 05:59:02 PM EST | | Running CPU benchmarks
Mon 16 Nov 2020 05:59:03 PM EST | | Suspending computation - CPU benchmarks in progress
Mon 16 Nov 2020 05:59:33 PM EST | | Benchmark results:
Mon 16 Nov 2020 05:59:33 PM EST | | Number of CPUs: 16
Mon 16 Nov 2020 05:59:33 PM EST | | 5840 floating point MIPS (Whetstone) per CPU
Mon 16 Nov 2020 05:59:33 PM EST | | 22791 integer MIPS (Dhrystone) per CPU
Mon 16 Nov 2020 05:59:34 PM EST | | Resuming computation

---

So I should be able to run four models at a time, if I ever get any. Right?

Easily. In fact the total throughput for a given amount of time will increase with more models, up to a point. It's just that once you go above 4, the time to completion for individual models starts increasing significantly. You could probably do 6 or 7 at a time and get maximum total work finished per unit time. You definitely do not want to run more than 8 as these do not do well when the number of models exceeds the number of physical cores.

I would love to run Linux in VM. Which flavour of Linux would be easy for someone to understand who has no experience of it? Easy to install and with a Windows type GUI.
I have twenty-four threads all dedicated to CPDN, sitting idle so might as well have a go at the Linux WU's.

Les has a good suggestion with Linux Mint. It is similar to Windows with its interface and based on Ubuntu, which the mods here run. Commands for installing the proper 32bit libraries are well documented.

There are different types of VM software you can run on Windows to run/manage the VMs. Like Alan K, I use VMWare Workstation Player as it's pretty easy to use. When installing within VMWare, you can specify the number of "processors" used, the amount of RAM to dedicate to the VM, etc. If you specify processors, say 4, then when running cpdn on boinc, I would limit the number of models to half that (in this case 2) so they aren't running on hyperthreads. Since it appears you are running on laptops, that may be enough anyway.

Something that may be important about Mint: It comes with several desktops.
I'm using Cinnamon, which gives it the Windows appearance.

And regarding "freezing" - I get that too.
I suspect that it's because I'm using all of the L3, and the OS wants some for itself at certain times.
It's just a matter of patience for a few seconds.

So I should be able to run four models at a time, if I ever get any. Right?


Easily. In fact the total throughput for a given amount of time will increase with more models, up to a point. It's just that once you go above 4, the time to completion for individual models starts increasing significantly. You could probably do 6 or 7 at a time and get maximum total work finished per unit time. You definitely do not want to run more than 8 as these do not do well when the number of models exceeds the number of physical cores.

I am thinking about that. Once I start getting ClimatePrediction work units, I may just reboot and diddle the BIOS to turn off hyperthreading. For my normal work, a core or two, or even four, should be enough;

---

It may be worth trying increasing the number of days work requested above the default. I know Les didn't find anything in work fetch debug that was helpful.

Do you have a suggestion as to what this should be? I know with different hardware and different OS, the amount will be different, but right now mine is set to

Store at least 0.3 days of work.
Store up to an additional 1 day of work.

Is this ridiculous? Or is it reasonable?

---