In the case of esxi, the kernel allocates the ram you've set in the vm properties up to the limit (options of the vm).
If resource contention starts, esxi tries to reclaim memory from guests by using the balloon driver. This is when you will see paging by the vm os to it's page file. If that process fails, esxi begins paging out physical ram of the vm to the vm host swap file (one per guest). First thing I would check on the vm guest is that there is no limit set. (I'm not at my desk right now, I forget what the tab is labeled). Make sure ram is set to unlimited so that the only ram limit is that specified on the first scren of the vm. Steven J. Sumichrast On Nov 13, 2011, at 12:07 PM, Phillip Burk <[email protected]> wrote: > Steven, thanks for the reply. I am completely new to virtualization but am a > systems admin by trade. I am going to check the logging levels on the host. > I do suspect that swapping is occurring on either/both the host and guest but > checking the guest in top shows minimal pageouts. > > I was in the process of trying to figure out how virtual memory works in ESXi > 4.1. That process left me feeling less than adequate as a sysadmin. :) > > On Nov 13, 2011, at 12:43 PM, Steven J. Sumichrast wrote: > >> That's quite odd. I'm running multiple tf2 servers all at 24 slots, almost >> always full. All are Debian 5.0.8 x32. Have been running them on esxi 4 and >> 4.1 and never heard any complaints. Are you having the engine do verbose >> logging to disk? I do have logging turned off, but that was a recent >> change. >> >> Are you swapping to disk a lot inside the vm? >> >> Steven J. Sumichrast >> >> On Nov 13, 2011, at 11:01 AM, Phillip Burk <[email protected]> wrote: >> >>> I have a TF2 server set up on a Debian 6 VM. Ping times to/from the server >>> from my laptop are quite reasonable: ~40 ms. I have no issues getting the >>> server to run but once more than one player joins performance begins to >>> degrade. The more players, the worse the performance. Last night we >>> tested the server with five players and experienced tons of lag, rubber >>> banding effects, and finally the server just froze. Except that the >>> server was still running and responsive to ssh. >>> >>> I checked the performance graphs in vSphere Client and noticed that to both >>> the datastore and virtual disk Read and Write were experiencing serious >>> latency that appeared to be happening when we were experiencing lag and >>> other issues in game. >>> >>> If anyone has any tips for tuning either the Debian VM or the CentOS host >>> that is running this VM I would greatly appreciate it. >>> >>> _______________________________________________ >>> To unsubscribe, edit your list preferences, or view the list archives, >>> please visit: >>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >> >> _______________________________________________ >> To unsubscribe, edit your list preferences, or view the list archives, >> please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, please > visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
