We have about 65 mod 3 for 112 servers (at the moment). About 1/2 of them are fat WAS virtual machines. Once we hit 50% utilization, we add more ...
Think of it this way... Would you rather have dedicated swap (real) disks to each virtual machine that may or may not be used or would you rather pool them all together and have VM's very robust block paging system manage full volumes and balance across all of them. Easier to monitor.. Your i/o load is spread... There's even multilevel storage of those pages (memory, xstor, disk). Marcy Cortes "This message may contain confidential and/or privileged information. If you are not the addressee or authorized to receive this for the addressee, you must not use, copy, disclose, or take any action based on this message or any information herein. If you have received this message in error, please advise the sender immediately by reply e-mail and delete this message. Thank you for your cooperation." -----Original Message----- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of James Melin Sent: Tuesday, December 04, 2007 6:20 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: [LINUX-390] Is 275GB of VDISK stupid? So Marcy, if you have 275GB of VDISK defined, what do you define for total page space for your VM? Just curious. I am almost always in disagreement with our VM guy on our configuration, but I'm getting ready to revisit VDISK swap because we're on a new machine with more memory. The more information I have on 'real companies' and their configurations the less intractable he seems to be. -J Marcy Cortes <[EMAIL PROTECTED]> Sent by: Linux on 390 Port <LINUX-390@VM.MARIST.EDU> To LINUX-390@VM.MARIST.EDU cc 12/03/2007 07:30 PM Subject Re: Is 275GB of VDISK stupid? Please respond to Linux on 390 Port <LINUX-390@VM.MARIST.EDU> I haven't added it all up, but we do it that way and would have to be either right around 275G if not over. One thing to look out for is the vm.swappiness setting. Seems that on SLES 9 (latest kernel) some virtual machines have a tendency to just keep writing more and more to swap and never free up those pages in vdisk until the process is stopped. We had production machines march through all their space and then java core dump :(. Used page space was also growing too quickly without it. Yes, swapoff/swapon would free up those pages (but the structures would still exist in VM). Setting kernel parameter vm.swappiness=20 seemed to prevent that. Don't ask me why... It just does :) (no time to go read the code and truly understand that)... The default is 60. PS. We use mod 3's for paging and mod 27's mostly for everything else. 9's seem kind of useless after you build a file system on them and only have like 6-something Gig left. Marcy Cortes ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
