[Xen-users] crazy load average on domU!

[Quartexx <quartex73@xxxxxxxxx>]

Hi.
this is the scenario:
dom0 xen 4.0.1 kernel 2.6.32.21.  dom0_mem=1024M,  dom0-min-mem 1024
enable-dom0-ballooning no
xen quad core, total ram 8GB

domU 1GB, 40GB VBD (tapdisk2), swap 4GB

VBDs are placed on a ext3 partition on top of drbd on top of md(raid1)
on top of a couple of sata disks

domU is running debian lenny with a lamp env (all officiale lenny
packages) hosting 1 site (it's not busy site) and 1 db (35MB).
After a while (sometimes some minutes or some hours) load increased
(today it reached 150!), all apache process goes in interruptable
sleep and system become unresponsive

vmstat shows:
 0 62 880188   8112    664  21128  492    4   492     4   82   96  0  0  0 100

If I'm fast enough and kill apache load decrease and system become
usable again.  If not then I have to destroy domU from dom0 and start
it again.
It looks like a I/O problem to me....

The important thing is that the same site with db works like a charm
on a physical machine (Intel Core 2 running hundreds of sites, a
vmware service with one guest running 2500!! sites, mail, ecc.).
Still I've tried tuning apache but with no luck.
So now I'm stuck because issue starts on xen 4.0 kernel 2.6.31 so I
upgraded to 4.0.1 and 2.6.32 but no luck.

When this domU becomes crazy, dom0 is ok (no high load, no swap, no
high cpu usage) and other 4 domU (one running nagios and cacti and 3
others running ipsec and openvpn) are good too.

There are a lot of factors I know.  maybe drbd on top of md? (but I
have some of these on non xen env and no issues).  kernel?  xen?

Any help will be well accepted. Thanks in advance

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users