Windows SMP Client, general troubleshooting

[ThunderRd]

Tried the double-VM setup after I saw it mentioned; it's an interesting idea but it was not stable. My results are OK in Linux on every WU I have had except 3060.

[ccperf721p]

Which distro are you on TR? Are the problems with both clients, or will one keep going after the other one fails..

[ThunderRd]

Running Ubuntu 7.04; have had multiple problems. Usually it's one client, and the other runs on. Have done all the standard stuff, static IP etc.

Have had a lot of core status 66 (102) which is simply a sigterm signal - the client just shuts down along with the window. It is recoverable with a restart, and sometimes it will run fine after that, although at other times it just repeatedly plotzes.

I also have had quite a few "long 1-4 interactions" errors. Most of these are also recoverable although some have actually hung the OS. I run a regular backup to restore these from if the OS crashes.

I can't put a finger on why it's only this WU. Those 2653's and others run like **** through a goose.

[ccperf721p]

Link me to your thread so I know what you have tried and I don't end up repeating everything that has already been said..

I'm sure you have already tried disabling -forceasm but it's always a good place to start..

[cw823]

Originally Posted by ThunderRd

Tried the double-VM setup after I saw it mentioned; it's an interesting idea but it was not stable. My results are OK in Linux on every WU I have had except 3060.

I have it on 5.....very easy and very stable. You couldn't have gotten a 3060 in a VM, so you must be changing the subject.

Were you setting affinity for the VMs? That is a must.

[ThunderRd]

Originally Posted by cw823

I have it on 5.....very easy and very stable. You couldn't have gotten a 3060 in a VM, so you must be changing the subject.

I'm not sure whether to read this as informative or critical. I would hope it is the first.

I can't control what I get, so I tried it with what I did get. It wasn't stable. I think it was a 2653. If you are referring to my posting etiquette, well...

Were you setting affinity for the VMs? That is a must.

No, and that's because I probably missed that bit in your instructions. Also, I'm not entirely sure how to do it in Linux; I'm still learning it. If you give me a procedure I'll be happy to try it again and see if it stabilizes for me. |

Ahh..... by chance do you mean to set the Windows affinities for the running VMs? I didn't think of that until just now. If that is the case, what is the process name, because there are several that VMWare runs.

[cw823]

Originally Posted by ThunderRd

I'm not sure whether to read this as informative or critical. I would hope it is the first.

I can't control what I get, so I tried it with what I did get. It wasn't stable. I think it was a 2653. If you are referring to my posting etiquette, well...

Unless you had horrible grammar or mispellings, I would not refer to your posting etiquette. You talked about VMs, which use 2 processors, then referred to a Quad WU, which sees (and uses) 4 processors so I assumed you had changed the subject. Apologies if that doesn't make sense

Originally Posted by ThunderRd

Ahh..... by chance do you mean to set the Windows affinities for the running VMs? I didn't think of that until just now. If that is the case, what is the process name, because there are several that VMWare runs.

Windows affinities, off the top of my head vmwarevmx or something like that.....once you're running FAH in the VM you'll see that particular file using approximately 40%. On a quad you want to set one VM to CPU1/2 and the other Vm to CPU3/4. Of course, should the VM get closed, once it is reopened you will need to manually set affinities again. Not a huge deal if it's a dedicated FAH machine, but can be a pain if it's a standalone.

Next chance I get I'll document the process start to finish.

All that I really know about linux is how to make it run FAH. Beyond that, I'm clueless.

[Daniel ~]

Have we got a how to on this? CW? hows about a how to? Great to see you back BTW! ":O}

[cw823]

Originally Posted by Daniel ~

Have we got a how to on this? CW? hows about a how to? Great to see you back BTW! ":O}

I have a Dell quad that I'm selling in 6 days (ebay), I'll throw another HD in it and write it up in the next couple days.

[Daniel ~]

Thank you very much!

[ThunderRd]

I understand the confusion. What I was saying is that in Linux, I seem to be able to fold any WU other than 3060. I don't have a lot of experience folding in VMs in windows, although I use VMs a lot to test software and operating systems. My experience in the last couple days was that I couldn't get the VMs stable on the WUs they got, which were 2653s, so I felt it was best to keep the machine running in Linux instead until I got more information. The problem is that each time Linux draws a 3060 I have problems, so that's the rub.

I'll take a look at setting the affinities in windows later and see if it does the trick for me.

Thanks.

[cw823]

I have all the screenshots now from start of VMWare install to setting FAH up as a service inside a VM (ubuntu). Just need to upload the pics, type up a post, and go. Hopefully tonight, but maybe tomorrow night.

[Daniel ~]

In the fullness of time...":O}

Looking forward to reading it!

[ccperf721p]

I was having some problems with some projects I am working on so I had to move back to a 32 bit linux install on both of my rigs. I've been folding single clients the last couple of days but will get back to the SMP clients in a Virtual Machine once I get this project finished up.

[ThunderRd]

On a happy note I spent some time poking around in the BIOS and the more obscure memory settings and managed to finish a 3060 from start to finish with no problems.

Now I'm on a pair of 3060s with 2 clients in Linux; one at 31% and one at 56%, everything looks stable.

Have not gone back to try the VMWare in windows but will try again soon.

[ThunderRd]

An important reminder for anyone running the Linux SMP client:

Folding Forum • View topic - [always use] -smp switch

I've done the same thing as this user once or twice. Don't be guilty yourselves. Like him, I usually use a script but on a couple of occasions have been poking around in the SMP directory and started it from there without the script- ouch! No getting it back unless you are running backups. I wasn't; but I am now...

[ThunderRd]

Looks like a new version of the Windows SMP client is coming soon, according to Vijay Pande:

Folding@home: New Windows client/core development (SMP and classic clients)

That's good news for guys having problems with the client. If they are using a new version of mpi I'm sure that will go a long way to getting the client more stable.

On another note, you may have noticed my non-existent production of late; I'm finally taking a long vacation and spending some time on the boat. I'm making a plan to sail to Malaysia, and then back to Krabi, where the boat is. That will most likely be next week, after I finish up some maintenance that's been due. I'll prolly be gone for the rest of the month, but I'll be checking in, although not as often as usual.

[Samuknow]

I so want to be you when I grow up....

Have a great time off.

[Daniel ~]

Have a great vacation, have an adventure! LOL! Be well!!

[ThunderRd]

If you remember the fracas caused by ccperf's thread a while back re: "dual linux smp clients on quads" then this thread will be of interest:

Folding Forum • View topic - 2619 What a PIA!

You will note that on page 2 of the thread our friend roadkill, aka 00john00, commences to get into a pissing contest with one of the moderators regarding his position on multiple clients and the payoff; the mod attempts to explain Pande Group's effort to get everyone using one client per core.

The thread is an interesting read and amusing at times, more so than others of its type in the past. The points vs. speed controversy has been legendary in the community forum in the last year. Personally, I can't wait for the new points system to get introduced so the arguments go away, but still, if you have any interest check it out and post your own take on it here...