linuxrouter

libertysyclone

i am having the following error when trying to run the VM? did I download the wrong VMplayer?

This is on the above mentioned DL 585 server

You need to grab version 3.0.0. The latest 3.0.1 is limited to 4-cores.

Speaking of which, why on earth would a new version only support 4 cores? How idiotic 

Governator

Speaking of which, why on earth would a new version only support 4 cores? How idiotic 

Yeah it does not make much sense to me. I was hoping to next release would increase the limit to 16, but instead they downgrade to 4.

linuxrouter

Governator

Speaking of which, why on earth would a new version only support 4 cores? How idiotic 

Yeah it does not make much sense to me. I was hoping to next release would increase the limit to 16, but instead they downgrade to 4.

It pretty much matches the direction of their main business - consolidating lots of small older servers on a newer faster system.  Their focus is really on 1 and 2 CPU VMs.  It figures they're not going to put a lot of extra effort into the free version.  We're really using it in a bass-ackwards way anyway - it would make more sense to run the CPU-intensive app (folding) natively and then the remaining work virtualized.

Sun's VirtualBox supports much larger SMP VMs.  I did an experiment this week on a dual Westmere system where I created a 12-core Linux VM and ran bigadv in it.  Frame times were about 5 times the native equivalent - 2hrs30min instead of 30min.  CPU time on the busy cores fluctuated between 20 and 90% every few seconds.  So, it seems like VirtualBox still has a long way to go.

Punchy

It pretty much matches the direction of their main business - consolidating lots of small older servers on a newer faster system.  Their focus is really on 1 and 2 CPU VMs.  It figures they're not going to put a lot of extra effort into the free version.  We're really using it in a bass-ackwards way anyway - it would make more sense to run the CPU-intensive app (folding) natively and then the remaining work virtualized.

Sun's VirtualBox supports much larger SMP VMs.  I did an experiment this week on a dual Westmere system where I created a 12-core Linux VM and ran bigadv in it.  Frame times were about 5 times the native equivalent - 2hrs30min instead of 30min.  CPU time on the busy cores fluctuated between 20 and 90% every few seconds.  So, it seems like VirtualBox still has a long way to go.

I keep hoping Virtualbox will continue to improve their SMP code. They have made some improvements. With their older versions, bigadv work units would crash almost instantly after startup. With the newer versions, I also notice the CPU fluctuation but the work unit seems to continue to run for the most part. With 8-threads, I see usage bounce around between 75 and 100%.

We'll see how these old procs do, they are 2005 AMD Opty's so i dont expect a lot from them.

Will adding more RAM into the VM help any? I see it currently using ~6GB (I have 16 available) 

Governator BTW you guys, I have my 980X stable at 4.3 GHz with 26:22 TPF 8-core smp I really should go native Linux for 12 cores and a probable 22 min TPF  but am concerned about having the two gpu clients working...

You know you want to switch... but yeah GPU under Linux is a big pain.
I got 22m15s per frame overnight on my x5670 running -smp 11 and it's only at 3.96GHz, so you might be under 21min at 4.3 = 50K ppd.

libertysyclone

We'll see how these old procs do, they are 2005 AMD Opty's so i dont expect a lot from them.

Will adding more RAM into the VM help any? I see it currently using ~6GB (I have 16 available) 

The default memory allocation should be enough. However, since you have the extra RAM available, you could set 5-6GB to the VM. This will give the Linux system a bit more breathing room. On average, I see around 450-500MB per thread of memory usage.

linuxrouter

Governator
Yes always when I do this sort of thing. I waited to complete a WU at the time, for it to be sent and to grab another and that's when I stooped it and edited. I restart and all it does is go to the prompt looking for my info every time. 

When you stopped the VM, did you enter the command reboot or shutdown? This command makes sure that the Linux system shuts down properly and the cache is flushed.

Boy you know, I just hit ctrl-c and then power down. I'll do the reboot next time all the time then :P

LR-check out the link that I posted on folding forum. Had the very same problem I posted here a while beck, client cannot upload completed WU and download the next. Could it be that the Decoupler is causing this issue?

http://foldingforum.org/v...c.php?f=18&t=14252

Punchy
If you do "ps -ef | grep lang" does langouste show up in the results?  If not, it may no longer be running.  However, since your log showed that downloads were at least being attempted, that doesn't seem to be the problem.

 
Have not done the langouste check yet. This is going on on one rig however my other one was able to get the next WU but is unable to upload the completed one. Last time this happened is I ended up losing the completed WU and ended having to restart a whole new image.

It's all good to go now. Picked up a new WU an hour ago. Thanks LR and Punchy.

I built and added the latest 2.6.34-rc5 kernel into the dev image. I am not sure yet if it performs any better since I have not had a chance to do a side-by-side comparison with 2.6.33. I would like to add it into the native image also to see if it scales any better than 2.6.33 with greater # of cores.

I have one bigadv on my 980X rig, started up another on my 920 machine tonight and it came up with the same IP, does this happen often? I'm guessing that this is not what we want obviously, how do you change IP on vmware anyway?

linuxrouter

Both should be getting unique IP addresses from the DHCP server. Did you setup a fresh copy of the image on the second system?

That's what I figured, it threw me when I saw that. By fresh, you mean recent version? I have 1.2 image on both.

linuxrouter

Governator
That's what I figured, it threw me when I saw that. By fresh, you mean recent version? I have 1.2 image on both. 

When you setup the image on the second system, was it directly from the original compressed file or did you copy the image from the first system and reconfigure?

Otherwise, you can try these commands out to refresh the DHCP in the VM:


/etc/rc.d/rc.inet stop
rm /etc/dhcpc/dhcpcd.duid

/etc/rc.d/rc.inet start
info

The info command will tell you the IP setup.

I used the compressed image 1.2 stored on an archive dvd. I'll try those commands if it happens again as I'm reinstalling vmware right now and will let you know.