[Openstack] iops limiting with OpenStack Nova using Ceph/Network Storage

Tyler Wilson kupo at linuxdigital.net
Wed Aug 27 16:42:52 UTC 2014


Thanks for the bug report, I'll keep an eye on the progress there. I've
attempted the patch mentioned and I can't seem to get it to work. Here is
my block under tune_items;

http://pastebin.mozilla.org/6181109

I've applied this to all nova-compute servers and restarted nova-compute
after the changed, spawned a new instance and didn't see any iotune blocks
in the domain's xml. Here is my current extra_specs for m1.small;
{"quota:write_iops_sec": "1024", "quota:disk_write_bytes_sec": "10240000",
"quota:vif_outbound_average": "10240", "quota:cpu_period": "10000",
"quota:disk_total_iops_sec": "10240", "quota:vif_inbound_average": "10240",
"quota:read_iops_sec": "1024", "quota:disk_write_iops_sec": "1024",
"quota:disk_read_bytes_sec": "10240000", "quota:cpu_quota": "10000",
"quota:disk_read_iops_sec": "1024", "quota:disk_total_bytes_sec": "10240"}


On Wed, Aug 27, 2014 at 4:19 AM, Yaguang Tang <heut2008 at gmail.com> wrote:

> CONF.libvirt.images_default_iops_second isn't in upstream yet,  anyway I
> think this is definitely a bug and we should support set rate limit for rbd
> disk backend.
> I filed a bug here to track this issue.
> https://bugs.launchpad.net/nova/+bug/1362129
>
>
>
> On Wed, Aug 27, 2014 at 7:05 PM, Haomai Wang <haomaiwang at gmail.com> wrote:
>
>> Yes, it's possible just like a volume from Cinder. But it still exists
>> some works todo. We need to pass iops/bw throttle value in via config
>> or flavor metadata. If you want to get it as quick as possible, you
>> can add some hack codes like this:
>>
>>         tune_items = ['disk_read_bytes_sec', 'disk_read_iops_sec',
>>             'disk_write_bytes_sec', 'disk_write_iops_sec',
>>             'disk_total_bytes_sec', 'disk_total_iops_sec']
>>         # Note(yaguang): Currently, the only tuning available is Block I/O
>>         # throttling for qemu.
>>         if self.source_type in ['file', 'block']:
>>             for key, value in extra_specs.iteritems():
>>                 scope = key.split(':')
>>                 if len(scope) > 1 and scope[0] == 'quota':
>>                     if scope[1] in tune_items:
>>                         setattr(info, scope[1], value)
>> +        if not getattr(info, 'disk_total_bytes_sec'):
>> +            setattr(info, 'disk_total_bytes_sec',
>> CONF.libvirt.images_default_bw_second)
>> +        if not getattr(info, 'disk_total_iops_sec'):
>> +            setattr(info, 'disk_total_iops_sec',
>> CONF.libvirt.images_default_iops_second)
>>         return info
>>
>> On Wed, Aug 27, 2014 at 4:32 PM, Tyler Wilson <kupo at linuxdigital.net>
>> wrote:
>> > Hey All,
>> >
>> > Is it possible to setup a iops/bytesps limitation within nova using
>> libvirt
>> > methods? I've found the following links but cant get it to work with my
>> > environment;
>> >
>> > http://ceph.com/planet/openstack-ceph-rbd-and-qos/
>> > https://wiki.openstack.org/wiki/InstanceResourceQuota
>> >
>> > I see in the commit code that it specifically mentions file and block
>> with
>> > no network in the code;
>> >
>> >         tune_items = ['disk_read_bytes_sec', 'disk_read_iops_sec',
>> >             'disk_write_bytes_sec', 'disk_write_iops_sec',
>> >             'disk_total_bytes_sec', 'disk_total_iops_sec']
>> >         # Note(yaguang): Currently, the only tuning available is Block
>> I/O
>> >         # throttling for qemu.
>> >         if self.source_type in ['file', 'block']:
>> >             for key, value in extra_specs.iteritems():
>> >                 scope = key.split(':')
>> >                 if len(scope) > 1 and scope[0] == 'quota':
>> >                     if scope[1] in tune_items:
>> >                         setattr(info, scope[1], value)
>> >         return info
>> >
>> > Is it possible to limit or establish QoS rules for network storage in
>> nova
>> > currently or only in cinder? My source protocol is rbd, qemu driver and
>> raw
>> > disk type.
>> >
>> > _______________________________________________
>> > 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
>> >
>>
>>
>>
>> --
>> Best Regards,
>>
>> Wheat
>>
>> _______________________________________________
>> 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
>>
>
>
>
> --
> Tang Yaguang
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140827/9eaf7a78/attachment.html>


More information about the Openstack mailing list