[Openstack] (no subject)

Adam Lawson alawson at aqorn.com
Fri Mar 28 01:29:03 UTC 2014


Swift is said to be eventually consistent because the data is stored then
eventually distributed in a balanced way. You don't need to manually
re-balance the rings constantly. Swift will do that for you. Re-balancing
rings is usually initiated after you *change the ring
structure*(add/remove regions, add/remove zones, change device
weights, etc).

In your case since you only have one node, Swift will distribute the
replicas across all 3 zones assuming you've configured 3x replication. When
you add a node and update the rings, yes you'll want to re-balance. That
will tell Swift to put a replica on the new node since Swift default
behavior is to keep replica placements "as unique as possible". That's the
actual Swift vernacular everyone uses. ; )

Unique replica placement strategy is as follows:

Region (if defined) > Zone > Node > Device > Device with fewest replicas


Good luck.

Adam


*Adam Lawson*
AQORN, Inc.
427 North Tatnall Street
Ste. 58461
Wilmington, Delaware 19801-2230
Toll-free: (888) 406-7620



On Thu, Mar 27, 2014 at 12:48 PM, Roman Kravets <softded at gmail.com> wrote:

> Dear Adam,
>
> I have one storage server and 12 hard drives on it.
> For test I split disk to 4 zones. If I rightly understood, swift load date
> during "re-balance ring" and load data right away to correct node.
>
> --
> Best regards,
> Roman Kravets
>
>
> On Thu, Mar 27, 2014 at 10:05 PM, Adam Lawson <alawson at aqorn.com> wrote:
>
>> Probably has to do with the fact you (I'm guessing) don't have very many
>> drives on that server. Is that a correct statement? I know that even with
>> 50 drives across a cluster (still very small), rings balance is at 100%
>> until the rings are adequately balanced. Look at your ring stats, drive
>> count and 5 zones for more consistent reports.
>>
>>
>> *Adam Lawson*
>> AQORN, Inc.
>> 427 North Tatnall Street
>> Ste. 58461
>> Wilmington, Delaware 19801-2230
>> Toll-free: (888) 406-7620
>>
>>
>>
>> On Thu, Mar 27, 2014 at 10:20 AM, Кравец Роман <softded at gmail.com> wrote:
>>
>>> Hello.
>>>
>>> I installed Openstack Swift to test server and upload 50 gb data.
>>> Now I see it in the log:
>>> root at storage1:/var/staff/softded# tail -n 1000 -f /var/log/syslog  |
>>> grep  replicated
>>> Mar 27 19:44:24 storage1 object-replicator 112746/187053 (60.27%)
>>> partitions replicated in 300.01s (375.81/sec, 3m remaining)
>>> Mar 27 19:47:44 storage1 object-replicator 187053/187053 (100.00%)
>>> partitions replicated in 499.71s (374.32/sec, 0s remaining)
>>> Mar 27 19:53:14 storage1 object-replicator 112863/187068 (60.33%)
>>> partitions replicated in 300.01s (376.20/sec, 3m remaining)
>>> Mar 27 19:56:29 storage1 object-replicator 187068/187068 (100.00%)
>>> partitions replicated in 494.53s (378.27/sec, 0s remaining)
>>> Mar 27 20:01:59 storage1 object-replicator 112343/187080 (60.05%)
>>> partitions replicated in 300.01s (374.47/sec, 3m remaining)
>>> Mar 27 20:05:18 storage1 object-replicator 187080/187080 (100.00%)
>>> partitions replicated in 498.55s (375.25/sec, 0s remaining)
>>> Mar 27 20:10:48 storage1 object-replicator 112417/187092 (60.09%)
>>> partitions replicated in 300.01s (374.71/sec, 3m remaining)
>>>
>>> Why object-replicator show different percent every time?
>>>
>>> Thank you!
>>>
>>> --
>>> Best regards,
>>> Roman Kravets
>>>
>>> _______________________________________________
>>> 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
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140327/b1d438f9/attachment.html>


More information about the Openstack mailing list