[mythtv-users] How much RAM for a 'Xen' backend-only virtual server
tfjellstrom at shaw.ca
Mon May 5 17:19:10 UTC 2008
On May 5, 2008, Gordon McLellan wrote:
> Yikes, 1g of ram is awful tight for a _server_ of any kind, let alone
> a virtualization host!
> If the xen host is "only" running the mythbackend (which also means
> it's running mysql), why does it need to be virtual? You mention
> you're already running mysql for something else, why can't the myth
> database be hosted on that instance too?
> I'm running mythbackend as a domu on my server, and I have allocated
> it 1g of ram and 2 vcpus, with Dom0 having 8g of ram and a quad core
> Xeon. Linux appears to be using most of that 1g as buffers and
> cache, but that is important, since disk io on a virtual machine, and
> disk io in general, is terribly slow, and a database should run from
> ram if at all possible.
> # free -m
> total used free shared buffers cached
> Mem: 1024 904 119 0 8 257
> -/+ buffers/cache: 638 385
> Swap: 509 0 509
> Running top and sorting by memory usage shows httpd as the biggest hog
> of ram, all ten of it's children claiming 45mb of ram each, the yum
> update daemon claiming 72mb, mythbackend claiming 39mb, and mysqld
> claiming 41mb. I forget how to tell how much of the db mysql has
> cached in memory, so I'll have to report back on that. In summary, I
> think my allocation of 1gb is excessive, but I've got the memory to
> waste on it, so what the hay.
> My setup worked perfectly for months and months with .20-1 (i think),
> but since upgrading to 20-2 and now 21, my uptime has been slashed to
> days at best. I don't know if there's more network IO occuring or
> what, but between the front and backends, I'm dealing with a lot of
> crashes related to networked file systems. I'm beginning to
> reconsider my use of virtualization at all, since I'm still having
> stability issues related to network storage problems. My myth
> network used to use NFS, and that would regularly crash, almost daily.
> After a few nfs crashes, it would leave the Dom0's kernel hosed and I
> would have to do a hard reset on the box. I've since switched to
> CIFS and enjoyed nearly a week of uptime before CIFS left Dom0's
> kernel hosed and again, another hard reset.
Huh, maybe its an issue with xen and nfs? I have been having the same
problems, my server would just crash, and most of the time the oopses came
from the vfs subsystem somewhere. I've since pulled the nfs share off the box
thats been giving me problems, and Its not crashed. And the box with the
share now is running vmware server instead, and seems to be doing ok.
As to why you might want the backend in a DomU, well any cpu use on the host
kernel will lag out the IO on the guests. With Xen you really _want_ to put
as much as possible into VMs, and leave the Dom0 as empty as possible.
> Hope this helps!
> On Mon, May 5, 2008 at 11:27 AM, James Buckley
> <xanium4332 at googlemail.com> wrote:
> > Hi all,
> > Just wondering the minimum amount of RAM I should dedicate to a mythtv
> > backend. As it's running in a Xen DomU, it'll be the only service
> > runninng (Gentoo Linux).
> > I only have 1G of RAM in the server, and it's acting as my router, mail
> > server, web server, mysql + postgres + ldap database server, so RAM's
> > starting to get tight.
> > Cheers, James
> > _______________________________________________
> > mythtv-users mailing list
> > mythtv-users at mythtv.org
> > http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
> mythtv-users mailing list
> mythtv-users at mythtv.org
tfjellstrom at shaw.ca
More information about the mythtv-users