Mailing List Archive: Xen: API

XCP 1.0 and openvswitch problem high memory

Index | Next | Previous | View Threaded

nyckopro at gmail

May 7, 2012, 5:35 AM

Post #1 of 4 (473 views)
Permalink

XCP 1.0 and openvswitch problem high memory

Hello, I have a problem with XCP 1.0. In recent weeks I saw very
resource openvswitch cpu and memory, no memory. At point (after a
while, less than a week) reaches the limit (512) and also begins to
swap out the point of leaving inoperative the real host and hence
their virtual machines. Everything returns to normal when I restart
the daemon (/ etc / init.d / openvswitch) but I can not be doing it
all the time. Someone can give me some tips to keep looking?

Leave some logs that may be useful:
#tail -f /var/log/openvswitch/ovs-vswitchd.log
May 07 09:28:36|47029|timeval|WARN|6 ms poll interval (0 ms user, 0 ms
system) is over 9 times the weighted mean interval 1 ms (3462238
samples)
May 07 09:28:36|47030|timeval|WARN|context switches: 0 voluntary, 3 involuntary
May 07 09:28:36|47031|coverage|INFO|Skipping details of duplicate
event coverage for hash=ba09b798 in epoch 3462238

If you need any more information do not hesitate to ask me because I
can not find the way to the solution and I'm several weeks into this.

PD: my english sucks

Regards
--
nycko

_______________________________________________
Xen-api mailing list
Xen-api [at] lists
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api

blp at cs

May 7, 2012, 12:35 PM

Post #2 of 4 (406 views)
Permalink

Re: XCP 1.0 and openvswitch problem high memory [In reply to]

Nycko <nyckopro-Re5JQEeQqe8AvxtiuMwx3w [at] public> writes:

> Hello, I have a problem with XCP 1.0. In recent weeks I saw very
> resource openvswitch cpu and memory, no memory. At point (after a
> while, less than a week) reaches the limit (512) and also begins to
> swap out the point of leaving inoperative the real host and hence
> their virtual machines. Everything returns to normal when I restart
> the daemon (/ etc / init.d / openvswitch) but I can not be doing it
> all the time. Someone can give me some tips to keep looking?

What version of OVS is this?

_______________________________________________
Xen-api mailing list
Xen-api [at] lists
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api

blp at cs

May 7, 2012, 1:16 PM

Post #3 of 4 (449 views)
Permalink

Re: XCP 1.0 and openvswitch problem high memory [In reply to]

Nycko <nyckopro [at] gmail> writes:

> On Mon, May 7, 2012 at 4:35 PM, Ben Pfaff <blp [at] cs> wrote:
>> Nycko <nyckopro-Re5JQEeQqe8AvxtiuMwx3w [at] public> writes:
>>
>>> Hello, I have a problem with XCP 1.0. In recent weeks I saw very
>>> resource openvswitch cpu and memory, no memory. At point (after a
>>> while, less than a week) reaches the limit (512) and also begins to
>>> swap out the point of leaving inoperative the real host and hence
>>> their virtual machines. Everything returns to normal when I restart
>>> the daemon (/ etc / init.d / openvswitch) but I can not be doing it
>>> all the time. Someone can give me some tips to keep looking?
>>
>> What version of OVS is this?
>
> ovs-vswitchd (Open vSwitch) 1.0.2
> Compiled Feb 7 2011 17:28:48
> OpenFlow versions 0x1:0x1

1.0.2 is an antique. I can't recommend that anyone use it any
longer.

I don't remember this specific problem, although it's almost
certainly a memory leak.

Presumably newer versions of XCP are using newer versions of OVS
as well. I hope so.

_______________________________________________
Xen-api mailing list
Xen-api [at] lists
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api

george.shuklin at gmail

May 7, 2012, 1:57 PM

Post #4 of 4 (462 views)
Permalink

Re: XCP 1.0 and openvswitch problem high memory [In reply to]

I've got that problem periodically (XCP 0.5 we still using for older
pool in product) in case of low resources for dom0. Solution: add more
cpu to dom0, raise memory for dom0.

After relaxing resource problem I've never get that condition for
prolonged time (>4 month) under pretty high loaded pool.

On 07.05.2012 16:35, Nycko wrote:
> Hello, I have a problem with XCP 1.0. In recent weeks I saw very
> resource openvswitch cpu and memory, no memory. At point (after a
> while, less than a week) reaches the limit (512) and also begins to
> swap out the point of leaving inoperative the real host and hence
> their virtual machines. Everything returns to normal when I restart
> the daemon (/ etc / init.d / openvswitch) but I can not be doing it
> all the time. Someone can give me some tips to keep looking?
>
> Leave some logs that may be useful:
> #tail -f /var/log/openvswitch/ovs-vswitchd.log
> May 07 09:28:36|47029|timeval|WARN|6 ms poll interval (0 ms user, 0 ms
> system) is over 9 times the weighted mean interval 1 ms (3462238
> samples)
> May 07 09:28:36|47030|timeval|WARN|context switches: 0 voluntary, 3 involuntary
> May 07 09:28:36|47031|coverage|INFO|Skipping details of duplicate
> event coverage for hash=ba09b798 in epoch 3462238
>
> If you need any more information do not hesitate to ask me because I
> can not find the way to the solution and I'm several weeks into this.
>
> PD: my english sucks
>
> Regards

_______________________________________________
Xen-api mailing list
Xen-api [at] lists
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads