[Gluster-users] Increasing replica count from 2 to 3

[Joe Julian]

Which application is filling memory?

If it's a brick (glusterfsd) then stopping and starting a brick ("kill" 
and "gluster volume start ... force") will not waste cycles re-healing 
any files that are healthy. Any heals of an individual file that were 
not complete will be restarted as well as any files that were changed 
while the brick was offline.

If it's glusterd, that can be restarted at any time without interfering 
with the volume.

If it's glustershd ("/usr/bin/glusterfs -s localhost --volfile-id 
gluster/glustershd ..."), you can restart that with "gluster volume 
start ... force" (even if the volume is already started).


On 12/29/2016 02:27 PM, Jackie Tung wrote:
> Ravi,
>
> Got it thanks.  I’ve kicked this off, it seems be doing OK.
>
> I am a little concerned about a slow creep of memory usage:
>
> * swap (64GB) completed filled up on server_1
> * general memory usage creeping up slowly over time.
>
> $ free -m
>             total        used        free      shared buff/cache   
> available
> Mem:         128829       55596         614          53 72618       71783
> Swap:         61034       61034           0
>
> Similar issue on server_2, though lower starting memory usage:
>
> The “available” number is slowly going down - at this rate, probably 
> will go to 0 before heal is done.
>
> We are actually running 3.8.6, I’d like to try to pause the heal, 
> upgrade to 3.8.7, and resume.  Is this possible heal suspend/resume 
> possible or advisable?
>
> The upgrade idea came from this on bugzilla (not 100% if it will help 
> my leak):
>
> https://bugzilla.redhat.com/show_bug.cgi?id=1400927
>
> Even without doing the upgrade, I may need to restart glusterfs-server 
> anyway to reset memory usage.
>
> Thanks,
> Jackie
>
>> On Dec 28, 2016, at 9:40 PM, Ravishankar N <ravishankar at redhat.com 
>> <mailto:ravishankar at redhat.com>> wrote:
>>
>> On 12/29/2016 10:46 AM, Jackie Tung wrote:
>>> Thanks very much for the advice.
>>>
>>> Would you mind elaborating on the "no io" recommendation?  It's 
>>> somewhat hard for me to guarantee this without a long maintenance 
>>> window.
>>>
>>> What is the consequence of having IO at point of add-brick, and for 
>>> the heal period afterwards?
>>
>> Sorry I wasn't clear. Since you're running  16 distribute legs 
>> (16x2), a lot of self-heals would be running and there is a chance 
>> that clients might experience slowness due to the self-heals. Other 
>> than that it should be fine.
>> Thanks,
>> Ravi
>>
>>>
>>>
>>>
>>> On Dec 28, 2016 8:27 PM, "Ravishankar N" <ravishankar at redhat.com 
>>> <mailto:ravishankar at redhat.com>> wrote:
>>>
>>>     On 12/29/2016 07:30 AM, Jackie Tung wrote:
>>>>     Version is 3.8.7 on Ubuntu xenial.
>>>>
>>>>     On Dec 28, 2016 5:56 PM, "Jackie Tung" <jackie at drive.ai
>>>>     <mailto:jackie at drive.ai>> wrote:
>>>>
>>>>         If someone has experience to share in this area, i'd be
>>>>         grateful.  I have an existing distributed replicated
>>>>         volume, 2x16.
>>>>
>>>>         We have a third server ready to go. Redhat docs say just
>>>>         run add brick replica 3, then run rebalance.
>>>>
>>>>         The rebalance step feels a bit off to me.  Isn't some kind
>>>>         of heal operation in order rather than rebalance?
>>>>
>>>>         No additional usable space will be introduced, only replica
>>>>         count increase from 2 to 3.
>>>>
>>>
>>>     You don't need to run re-balance for increasing the replica
>>>     count. Heals should automatically be triggered when you run
>>>     'gluster vol add-brick <volname> replica 3 <list of bricks for
>>>     the 3rd replica>`. It is advisable to do this when there is no
>>>     I/O happening  on the volume. You can verify that files are
>>>     getting populated in the newly added bricks post running the
>>>     command.
>>>
>>>     -Ravi
>>>>
>>>>
>>>>         Thanks
>>>>         Jackie
>>>>
>>>>
>>>>     The information in this email is confidential and may be
>>>>     legally privileged. It is intended solely for the addressee.
>>>>     Access to this email by anyone else is unauthorized. If you are
>>>>     not the intended recipient, any disclosure, copying,
>>>>     distribution or any action taken or omitted to be taken in
>>>>     reliance on it, is prohibited and may be unlawful.
>>>>
>>>>
>>>>
>>>>     _______________________________________________
>>>>     Gluster-users mailing list
>>>>     Gluster-users at gluster.org <mailto:Gluster-users at gluster.org>
>>>>     http://www.gluster.org/mailman/listinfo/gluster-users
>>>>     <http://www.gluster.org/mailman/listinfo/gluster-users>
>>>
>>> The information in this email is confidential and may be legally 
>>> privileged. It is intended solely for the addressee. Access to this 
>>> email by anyone else is unauthorized. If you are not the intended 
>>> recipient, any disclosure, copying, distribution or any action taken 
>>> or omitted to be taken in reliance on it, is prohibited and may be 
>>> unlawful.
>>>
> The information in this email is confidential and may be legally 
> privileged. It is intended solely for the addressee. Access to this 
> email by anyone else is unauthorized. If you are not the intended 
> recipient, any disclosure, copying, distribution or any action taken 
> or omitted to be taken in reliance on it, is prohibited and may be 
> unlawful.
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20161229/97b11490/attachment.html>