cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nux! <...@li.nux.ro>
Subject Re: Regular total loss of connectivity
Date Tue, 02 Jun 2015 12:52:51 GMT
Ok, managed to find the problem. It's a combination of settings on the provider cisco port
and the VR.
I'm not sure how this kind of problems always find me, perhaps some gypsy curse. :-)

The public IPs I am using are statically assigned in the cisco vlan and some random private
IP is assigned to this port so that it stays up.
Because this private IP is not part of the subnet I am using and also because arp_ignore is
set to 2 in the VR, all connectivity is lost after 14400 seconds (default arp cache in cisco,
I am told).

Problem can be fixed by either changing arp_ignore to 0 (what I did) or set a proper IP from
the same subnet on the cisco port.

What makes me scratch my head is why this problem was triggered all of a sudden ... it worked
fine for many months before the power cut.

Thanks

--
Sent from the Delta quadrant using Borg technology!

Nux!
www.nux.ro

----- Original Message -----
> From: "Nux!" <nux@li.nux.ro>
> To: dev@cloudstack.apache.org
> Cc: users@cloudstack.apache.org
> Sent: Monday, 1 June, 2015 22:19:58
> Subject: Re: Regular total loss of connectivity

> I was using latest stock EL6 , but now am trying kernel 4.x and qemu-rhev from
> ovirt.org.
> 
> Thanks,
> Lucian
> 
> --
> Sent from the Delta quadrant using Borg technology!
> 
> Nux!
> www.nux.ro
> 
> ----- Original Message -----
>> From: "Rafael Fonseca" <rsafonseca@gmail.com>
>> To: dev@cloudstack.apache.org
>> Cc: users@cloudstack.apache.org
>> Sent: Monday, 1 June, 2015 18:42:23
>> Subject: Re: Regular total loss of connectivity
> 
>> There are also a few kvm bugs which could be related, what kernel and
>> qemu-kvm versions are you running?
>> 
>> Rafael
>> 
>> On Mon, Jun 1, 2015 at 7:23 PM, Simon Weller <sweller@ena.com> wrote:
>> 
>>> What does arp show you on the VR when this occurs?
>>> Can you isolate this VR to a different physical host and upstream switch?
>>> If you leave a ping going to some external ip (e.g. 8.8.8.8) from the VR,
>>> do you still lose connectivity?
>>>
>>> ________________________________________
>>> From: Nux! <nux@li.nux.ro>
>>> Sent: Monday, June 1, 2015 12:15 PM
>>> To: users@cloudstack.apache.org
>>> Cc: dev@cloudstack.apache.org
>>> Subject: Re: Regular total loss of connectivity
>>>
>>> Ok, no luck with e1000 either, connectivity is lost after 4 hours (14400
>>> sec). That can't be random.
>>>
>>> Any ideas?
>>>
>>> --
>>> Sent from the Delta quadrant using Borg technology!
>>>
>>> Nux!
>>> www.nux.ro
>>>
>>> ----- Original Message -----
>>> > From: "Nux!" <nux@li.nux.ro>
>>> > To: users@cloudstack.apache.org
>>> > Cc: dev@cloudstack.apache.org
>>> > Sent: Monday, 1 June, 2015 14:35:09
>>> > Subject: Re: Regular total loss of connectivity
>>>
>>> > Nope, it's a regular, non-redundant VR.
>>> >
>>> > I've switched to using e1000 instead of virtio, waiting for a few hours,
>>> so how
>>> > it pans out. :-)
>>> >
>>> > --
>>> > Sent from the Delta quadrant using Borg technology!
>>> >
>>> > Nux!
>>> > www.nux.ro
>>> >
>>> > ----- Original Message -----
>>> >> From: "Simon Weller" <sweller@ena.com>
>>> >> To: dev@cloudstack.apache.org
>>> >> Cc: "Cloudstack Users List" <users@cloudstack.apache.org>
>>> >> Sent: Monday, 1 June, 2015 13:32:26
>>> >> Subject: Re: Regular total loss of connectivity
>>> >
>>> >> Is this VR in a redundant pair? If so, does stopping the master and
>>> allowing the
>>> >> slave to take over allow the flow of traffic to resume?
>>> >>
>>> >>
>>> >> ________________________________________
>>> >> From: Nux! <nux@li.nux.ro>
>>> >> Sent: Monday, June 1, 2015 6:45 AM
>>> >> To: dev@cloudstack.apache.org
>>> >> Cc: Cloudstack Users List
>>> >> Subject: Re: Regular total loss of connectivity
>>> >>
>>> >> Thanks Simon,
>>> >>
>>> >> link up/down has not helped, setting tso etc off on the link has not
>>> helped
>>> >> either.
>>> >> Connectivity is lost as usual after ~4 hours.
>>> >>
>>> >> I found some suggestions to try and use the e1000 nic instead of
>>> virtio, will do
>>> >> that.
>>> >>
>>> >> Lucian
>>> >>
>>> >> --
>>> >> Sent from the Delta quadrant using Borg technology!
>>> >>
>>> >> Nux!
>>> >> www.nux.ro
>>> >>
>>> >> ----- Original Message -----
>>> >>> From: "Simon Weller" <sweller@ena.com>
>>> >>> To: "dev" <dev@cloudstack.apache.org>, "Cloudstack Users List"
>>> >>> <users@cloudstack.apache.org>
>>> >>> Sent: Sunday, 31 May, 2015 22:36:56
>>> >>> Subject: Re: Regular total loss of connectivity
>>> >>
>>> >>> If you ifdown the interface on the router and then ifup it again,
does
>>> the arp
>>> >>> problem resolve itself?
>>> >>>
>>> >>> We've seen a similar issue before caused by malicious/heavy traffic
>>> related to
>>> >>> this bug:
>>> >>>
>>> >>> https://bugs.launchpad.net/ubuntu/+source/qemu-kvm/+bug/997978
>>> >>>
>>> >>> - Si
>>> >>> ________________________________________
>>> >>> From: Nux! <nux@li.nux.ro>
>>> >>> Sent: Sunday, May 31, 2015 4:25 PM
>>> >>> To: dev; Cloudstack Users List
>>> >>> Subject: Regular total loss of connectivity
>>> >>>
>>> >>> Hi,
>>> >>>
>>> >>> Following a power cut, one of my cloudstack deployments is having
a
>>> really weird
>>> >>> problem that I cannot seem to solve on my own.
>>> >>> Every 3 hours all the public IPs on the VR stop responding from
the
>>> Internet.
>>> >>> From the VR they are of course all reachable.
>>> >>> In the same VLAN as the public IPs there is another physical server,
>>> this one
>>> >>> can also access the VMs on their IPs just fine.
>>> >>>
>>> >>> The provider has not found the problem and hints at problems with
the
>>> cloud
>>> >>> platform, however cloudstack worked just fine until the power cut,
not
>>> to
>>> >>> mention the problem persists through HV and ACS upgrades.
>>> >>>
>>> >>> I'm thinking network side arp issues or something like this, alas
I am
>>> not that
>>> >>> good with network stuff and don't have access to it anyway.
>>> >>>
>>> >>> If I reboot the VR once or twice the IPs start working again and
the
>>> VMs are
>>> >>> accessible from the internet.
>>> >>>
>>> >>> Ideas?
>>> >>>
>>> >>> Env: CentOS 6, KVM, ACS 44 to 4.5.1, Adv zone
>>> >>>
>>> >>> Lucian
>>> >>>
>>> >>> --
>>> >>> Sent from the Delta quadrant using Borg technology!
>>> >>>
>>> >>> Nux!
> >> > > > www.nux.ro

Mime
View raw message