[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-users] strange xen memory calculations?
Hi, "trilok nuwal" <tc.nuwal@xxxxxxxxx> wrote: > It looks like memory balooning issue. > You should assigning the memory in the multiple of 256MB instead of any > random no. > It could be useful. I can set the dom0_mem to a multiple of 256, but not the rest. I use heartbeat and the Xen resource script there, to compute the optimal memory for the amount of running machines, sth. like this: domU_mem = total_memory - dom0_mem / count(domU's) So it just shouldn't be necessary to ballon, as the HA resource script calculates it before starting the machine. Right now it seems that is has to do with the amount of network bridges I create. I have two network cards bonded together, and a third single eth2 card. One bridge is on the single card, just for one of the virtual nodes. On top of the bond interface, I have 10 vlans, on each of the vlans I created a bridge. then in the domU config it looks like this: vif=[ 'mac=00:16:3e:14:50:26,bridge=bridge10', 'mac=00:16:3e:14:51:26,bridge=bridge1', 'mac=00:16:3e:14:52:26,bridge=bridge3', 'mac=00:16:3e:14:53:26,bridge=bridge4', 'mac=00:16:3e:14:54:26,bridge=bridge5', 'mac=00:16:3e:14:55:26,bridge=bridge6', 'mac=00:16:3e:14:56:26,bridge=bridge7', 'mac=00:16:3e:14:57:26,bridge=bridge8', 'mac=00:16:3e:14:58:26,bridge=bridge9', 'mac=00:16:3e:14:59:26,bridge=bridge12', ] I just made the test, removed all additional network bridges, and with only one vif configured in the domU's it is working as expected. I have the heartbeat managing xen instances in another cluster this way, without a problem. There I only have one virtual interface per virtual node. This is also sles10sp1. I had a opensuse 10.3. with xen 3.1. running with these multiple virtual interface using heartbeat and xen, and there it also was working fine. There I have the problem with the bonding interface, when I enable bonding, and do e.g. a lot of nfs traffic over it, the 10.3. is freezing completely. That was the reason to go by a sles10. As I am able now to reproduce the issue, I created a bug report at Novell bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=356325 Is there an options that limits the amount of virtual interfaces per xen instance that I can raise, or any other idea how I could make it work on the sles10sp1? thanks Sebastian > > Thanks, > Trilok > > On Jan 25, 2008 6:02 PM, Sebastian Reitenbach <sebastia@xxxxxxxxxxxxxxxxxxxx> > wrote: > > > Hi, > > > > I have started the xen kernel with parameter dom0_mem=300M. > > > > xm info | grep total > > total_memory : 2045 > > > > Name ID Mem VCPUs State Time(s) > > Domain-0 0 300 2 r----- 292.1 > > FTP 822 1 56.7 > > LDAP 1 822 1 r----- 541.3 > > NFS1 822 1 14.8 > > NFS2 822 1 1.4 > > > > For the domU's I have 822 mb configured, the first runs, when I want to > > start a second domU, then xend bails out because of missing memory: > > > > xm create /etc/xen/vm/NFSPUBLIC > > Using config file "/etc/xen/vm/NFSPUBLIC". > > Error: I need 841728 KiB, but dom0_min_mem is 65536 and shrinking to 65536 > > KiB would leave only 302336 KiB free. > > > > I know for all four domU's I do not have enough memory, but for two, it > > should work: > > > > 300 + 2*822 = 1944 < 2045 > > I don't understand why xend thinks it needs the memory, how is this > > calculated? I also set dom0_min_mem to 0 to stop ballooning at all, but > > that > > did not helped either. > > > > I am on sles10sp1 x86_64, using xen-3.0.4, the version that comes with the > > system. > > > > any hint is highly appreciated. > > Sebastian > > > > > > _______________________________________________ > > Xen-users mailing list > > Xen-users@xxxxxxxxxxxxxxxxxxx > > http://lists.xensource.com/xen-users > > > _______________________________________________ Xen-users mailing list Xen-users@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-users
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |