[Gluster-users] Maximising Heal Speed

[Pranith Kumar Karampuri]

On 04/14/2016 05:58 PM, Serkan Çoban wrote:
> Hi Pranith,
>
> Sure I can test it but not on physical nodes, doing tests on VMs ok for you?
> But at the end I can upgrade to 3.7.13 and do tests on physical nodes too.
It is not a good idea to do performance tests on VMs because there will 
be other things in the setup which affect performance :-(.

Pranith
>
> On Thu, Apr 14, 2016 at 2:04 PM, Pranith Kumar Karampuri
> <pkarampu at redhat.com> wrote:
>>
>> On 04/14/2016 04:15 PM, Serkan Çoban wrote:
>>>>>> We are almost done backporting multi-threaded self-heal. I am in the
>>>>>> process of writing a blog-post to give full idea about this one.
>>> Will multi thread self heal work for disperse volumes? what is the blog
>>> address?
>>
>> hi Serkan,
>>       At the moment we are enabling multi-threaded self-heal for only
>> replicate volumes, mainly because we didn't get enough time to test the
>> changes with disperse volumes. Do you want to help out with testing
>> multi-threaded heal in disperse volumes? I can provide a patch which does
>> this. We can may be target this for 3.7.13 based on your inputs?
>>
>> Pranith
>>
>>> On Thu, Apr 14, 2016 at 12:44 PM, Pranith Kumar Karampuri
>>> <pkarampu at redhat.com> wrote:
>>>>
>>>> On 04/14/2016 09:46 AM, Lindsay Mathieson wrote:
>>>>> Sorry to bring this up again, but I never did figure out the right
>>>>> settings for this.
>>>>>
>>>>> If I reboot a gluster node for a rep 3 volume, what are the settings
>>>>> for maximising heal speed, assuming ones not worried about i/o or cpu
>>>>>
>>>>> Gluster 3.7.9
>>>>> Sharded volume (4MB)
>>>> The only benefit you get over replicate volumes without sharding is that
>>>> only the shards which changed when the brick rebooted are healed instead
>>>> of
>>>> whole VM image. Apart from that I don't see any settings that will
>>>> improve
>>>> heal speed. You wait for 3.7.12. We are almost done backporting
>>>> multi-threaded self-heal. I am in the process of writing a blog-post to
>>>> give
>>>> full idea about this one.
>>>>
>>>> Pranith
>>>>>
>>>>> thanks,
>>>>>
>>>> _______________________________________________
>>>> Gluster-users mailing list
>>>> Gluster-users at gluster.org
>>>> http://www.gluster.org/mailman/listinfo/gluster-users
>>