[Openstack] nova-sheduler timeouts in juno ...

Jonathan Proulx jon at jonproulx.com
Thu Feb 12 20:59:00 UTC 2015


On Sun, Feb 8, 2015 at 12:14 PM, Tyler Wilson <kupo at linuxdigital.net> wrote:
> Try adjusting url_timeout in nova.conf, this is a new setting to allow more
> time to connect to neutron.

I increased the url_timeout from 30 to 120 this seems to have no
effect, still failing at 30 instances and in much less than 120sec.
So I don't think we're getting so far as trying to talk to neutron
yet.

Thanks,
-Jon


> On Sun, Feb 8, 2015 at 9:37 AM, Xinyuan Huang <huangxy0101 at gmail.com> wrote:
>>
>> Hi Jon,
>>
>> Could you please also paste the messages in nova scheduler logs, if
>> possible?
>>
>> It is true that in icehouse instances might get launched individually
>> before scheduler realizes no enough hosts are available, but currently this
>> has been changed and number of available hosts are checked before launching
>> any.
>>
>> Thanks,
>> Xinyuan
>>
>>
>> From: Jonathan Proulx <jon at jonproulx.com>
>> Date: Fri, Feb 6, 2015 at 7:25 AM
>> Subject: [Openstack] nova-sheduler timeouts in juno ...
>> To: "openstack at lists.openstack.org" <openstack at lists.openstack.org>
>>
>> >
>> >
>> > Hi All,
>> >
>> > After upgrading from icehouse to juno I get timeouts when trying to
>> > schedule >20 instances at once (20 deterministically works, 30 always
>> > fails didn't bother to go finer than that)
>> >
>> > Note this is a single call with --max-count on the CLI or setting
>> > number of instances in Horizon.  If I do parallel batched of 10 or 20
>> > instances I can get hundreds to launch successfully as each set gets
>> > handled by a different scheduler process.
>> >
>> > I know under icehouse I could get 100 or so to launch at once (then
>> > neutron would start to trip over itself, so kudos on that not
>> > happening anymore).  I seem to recall with icehouse instances in a
>> > scheduling batch would launch individually as scheduled in juno it
>> > seems none get launched until all are scheduled.
>> >
>> > In any case timeouts appear in nova-conductor.log one set for each
>> > instance in the set all with the same message ID
>> > (http://paste.openstack.org/show/168157/).
>> >
>> > Is anyone else seeing this / know a work around?
>> >
>> > Thanks,
>> > -Jon
>> >
>> > _______________________________________________
>> > Mailing list:
>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> > Post to     : openstack at lists.openstack.org
>> > Unsubscribe :
>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> >
>> >
>>
>> _______________________________________________
>> Mailing list:
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> Post to     : openstack at lists.openstack.org
>> Unsubscribe :
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>>
>




More information about the Openstack mailing list