I partially solve the problem.

My host mac=
hine has 2 network interfaces with a bond. The bond was 
configured wi=
th  mode=3D4 (802.3ad) and switch was configured in the same way.
If I=
 remove one network cable the network become stable. With both 
cables=
 attached the network is instable.

I removed the link aggregatio=
n configuration from switch and change the 
bond in mode=3D2 (balance-=
xor). Now the network are stable.
The strange thing is that previous c=
onfiguration worked fine for one 
year... since the last upgrade.

Now ha-agent don't reboot the hosted-engine anymore, but I receive t=
wo 
emails from brocker evere 2/5 minutes.
First a mail with "ovi=
rt-hosted-engine state transition 
StartState-ReinitializeFSM" and aft=
er "ovirt-hosted-engine state 
transition ReinitializeFSM-EngineStarti=
ng"

Il 17/12/2015 10.51, Stefano Danzi ha scritto:
 Hello,
 I have one testing host (o=
nly one host) with self hosted engine and 2 
 VM (one linux and one wi=
ndows).

 After upgrade ovirt from 3.6.0 to 3.6.1 the network con=
nection works 
 discontinuously.
 Every 10 minutes HA agent resta=
rt hosted engine VM because result 
 down. But the machine is UP,
 only the network stop to work for some minutes.
 I activate global m=
aintenace mode to prevent engine reboot. If I ssh 
 to the hosted engi=
ne sometimes
 the connection work and sometimes no.  Using VNC connect=
ion to engine 
 I see that sometime VM reach external network
 an=
d sometimes no.
 If I do a tcpdump on phisical ethernet interface I do=
n't see any 
 packet when network on vm don't work.

 Same t=
hing happens fo others two VM.

 Before the upgrade I never had n=
etwork problems.

 Users mailing list
 Users(a)ovirt.org
 http://lists.ovirt.org/mailman/lis=
tinfo/users