Re: Kernel Out of Memory error message
Ahh, the OOM killer. :) I typically only run in to this error when students accidentally put a fork in a loop and create a "fork bomb". In *nix, you can reduce the chance of this happening by putting a limit in the /etc/security/limits.conf, like: * hardnproc 200 # adjust this number to taste root - # make sure root can still have no limit That way no single user can have over (say) 200 processes, the second line allows root to go unlimited. You can also put some limits on RAM, etc. Some of these settings may be helpful. Having a limit can be bad, but not as bad as having the OOM killer come out on the loose. Once the grim reaper comes out bad things happen. :) John Jones East Carolina University College of Technology and Computer Science Science and Technology Bldg. Suite 100 Greenville, NC 27858-4353 252-737-1940, Fax 252-328-4250 -Original Message- From: Duhon, Gregory D Reply-to: "vcl-user@incubator.apache.org" To: 'vcl-user@incubator.apache.org' Cc: Olsen, Chris , Lee Thompson , Tadesse, Berhanu Subject: Kernel Out of Memory error message Date: Wed, 6 Oct 2010 13:30:14 -0400 Have any of you experienced this error message within your VCL environment? If so, how did you solve it? Oct 5 19:38:05 vclesx02 kernel: Out of memory: Killed process 28055, UID 0, (vmware-vmx). We currently have a minimum of 20 vmguest to a maximum of 30 vmguest allocated per blade within our VCL environment. With 48GB of memory per blade, we thought this was a ‘no-brainer’. Well, it isn’t. How many vmguest are most of you running per blade/computer within your VCL environment? Thanks, Gregory Duhon Infrastructure Architect Information Technology California State University, Northridge 818-677-4114 gregory.du...@csun.edu
RE: Kernel Out of Memory error message
Thanks, Jim. -Original Message- From: James O'Dell [mailto:jod...@fullerton.edu] Sent: Wednesday, October 06, 2010 11:28 AM To: vcl-user@incubator.apache.org Subject: Re: Kernel Out of Memory error message -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 This looks the Linux kernel "OOM Killer" killed vmware-vmx. Here's a nice article that describes it. > http://lwn.net/Articles/104179 Turning it off with a kernel setting could avoid it. But it may cause the system to just crash instead. You could add more swap space. But the system will be a really slow when it runs out of ram. I suggest avoiding over committing the ram by watching the memory size of the vm images. The vmware consultant we had help us virtualize our data center said you can over commit the CPU by about a factor of 2. But we should avoid(and monitor) over committing the RAM. Just my 2 cents, __Jim On 10/6/2010 10:30 AM, Duhon, Gregory D wrote: > Have any of you experienced this error message within your VCL > environment? If so, how did you solve it? > > > > Oct 5 19:38:05 vclesx02 kernel: Out of memory: Killed process 28055, > UID 0, (vmware-vmx). > > > > We currently have a minimum of 20 vmguest to a maximum of 30 vmguest > allocated per blade within our VCL environment. With 48GB of memory per > blade, we thought this was a 'no-brainer'. Well, it isn't. > > > > How many vmguest are most of you running per blade/computer within your > VCL environment? > > > > > > Thanks, > > Gregory Duhon > > Infrastructure Architect > > Information Technology > > California State University, Northridge > > 818-677-4114 > > gregory.du...@csun.edu > > > - -- Jim O'Dell Network Analyst California State University Fullerton Email: jod...@fullerton.edu Phone: (657) 278-2256 -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkysv6YACgkQREVHAOnXPYS2xgCZAebzYfbHmWMHr+28z6DN8T91 JQQAnR2I4kBHJ9oqh3bbDJul+ab7Ae/7 =sySg -END PGP SIGNATURE-
Re: Kernel Out of Memory error message
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 This looks the Linux kernel "OOM Killer" killed vmware-vmx. Here's a nice article that describes it. > http://lwn.net/Articles/104179 Turning it off with a kernel setting could avoid it. But it may cause the system to just crash instead. You could add more swap space. But the system will be a really slow when it runs out of ram. I suggest avoiding over committing the ram by watching the memory size of the vm images. The vmware consultant we had help us virtualize our data center said you can over commit the CPU by about a factor of 2. But we should avoid(and monitor) over committing the RAM. Just my 2 cents, __Jim On 10/6/2010 10:30 AM, Duhon, Gregory D wrote: > Have any of you experienced this error message within your VCL > environment? If so, how did you solve it? > > > > Oct 5 19:38:05 vclesx02 kernel: Out of memory: Killed process 28055, > UID 0, (vmware-vmx). > > > > We currently have a minimum of 20 vmguest to a maximum of 30 vmguest > allocated per blade within our VCL environment. With 48GB of memory per > blade, we thought this was a ‘no-brainer’. Well, it isn’t. > > > > How many vmguest are most of you running per blade/computer within your > VCL environment? > > > > > > Thanks, > > Gregory Duhon > > Infrastructure Architect > > Information Technology > > California State University, Northridge > > 818-677-4114 > > gregory.du...@csun.edu > > > - -- Jim O'Dell Network Analyst California State University Fullerton Email: jod...@fullerton.edu Phone: (657) 278-2256 -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.9 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkysv6YACgkQREVHAOnXPYS2xgCZAebzYfbHmWMHr+28z6DN8T91 JQQAnR2I4kBHJ9oqh3bbDJul+ab7Ae/7 =sySg -END PGP SIGNATURE-