[Gluster-users] Turn off replication
Karthik Subrahmanya
ksubrahm at redhat.com
Thu Apr 12 06:11:50 UTC 2018
On Wed, Apr 11, 2018 at 7:38 PM, Jose Sanchez <josesanc at carc.unm.edu> wrote:
> Hi Karthik
>
> Looking at the information you have provided me, I would like to make sure
> that I’m running the right commands.
>
> 1. gluster volume heal scratch info
>
If the count is non zero, trigger the heal and wait for heal info count to
become zero.
> 2. gluster volume remove-brick scratch *replica 1 *
> gluster02ib:/gdata/brick1/scratch gluster02ib:/gdata/brick2/scratch force
>
3. gluster volume add-brick* “#"* scratch gluster02ib:/gdata/brick1/
> scratch gluster02ib:/gdata/brick2/scratch
>
>
> Based on the configuration I have, Brick 1 from Node A and B are tide
> together and Brick 2 from Node A and B are also tide together. Looking at
> your remove command (step #2), it seems that you want me to remove Brick 1
> and 2 from Node B (gluster02ib). is that correct? I thought the data was
> distributed in bricks 1 between nodes A and B) and duplicated on Bricks 2
> (node A and B).
>
Data is duplicated between bricks 1 of nodes A & B and bricks 2 of nodes A
& B and data is distributed between these two pairs.
You need not always remove the bricks 1 & 2 from node B itself. The idea
here is to keep one copy from both the replica pairs.
>
> Also when I add the bricks back to gluster, do I need to specify if it is
> distributed or replicated?? and Do i need a configuration #?? for example
> on your command (Step #2) you have “replica 1” when remove bricks, do I
> need to do the same when adding the nodes back ?
>
No. You just need to erase the data on those bricks and add those bricks
back to the volume. The previous remove-brick command will make the volume
plain distribute. Then simply adding the bricks without specifying any "#"
will expand the volume as a plain distribute volue.
>
> Im planning on moving with this changes in few days. At this point each
> brick has 14tb and adding bricks 1 from node A and B, i have a total of
> 28tb, After doing all the process, (removing and adding bricks) I should be
> able to see a total of 56Tb right ?
>
Yes after all these you will have 56TB in total.
After adding the bricks, do volume rebalance, so that the data which were
present previously, will be moved to the correct bricks.
HTH,
Karthik
>
> Thanks
>
> Jose
>
>
>
>
> ---------------------------------
> Jose Sanchez
> Systems/Network Analyst 1
> Center of Advanced Research Computing
> 1601 Central Ave
> <https://maps.google.com/?q=1601+Central+Ave&entry=gmail&source=g>.
> MSC 01 1190
> Albuquerque, NM 87131-0001
> carc.unm.edu
> 575.636.4232
>
> On Apr 7, 2018, at 8:29 AM, Karthik Subrahmanya <ksubrahm at redhat.com>
> wrote:
>
> Hi Jose,
>
> Thanks for providing the volume info. You have 2 subvolumes. Data is
> replicated within the bricks of that subvolumes.
> First one consisting of Node A's brick1 & Node B's brick1 and the second
> one consisting of Node A's brick2 and Node B's brick2.
> You don't have the same data on all the 4 bricks. Data are distributed
> between these two subvolumes.
> To remove the replica you can use the command
> gluster volume remove-brick scratch replica 1 gluster02ib:/gdata/brick1/
> scratch gluster02ib:/gdata/brick2/scratch force
> So you will have one copy of data present from both the distributes.
> Before doing this make sure "gluster volume heal scratch info" value is
> zero. So copies you retain will have the correct data.
> After the remove-brick erase the data from the backend.
> Then you can expand the volume by following the steps at [1].
>
> [1] https://docs.gluster.org/en/latest/Administrator%
> 20Guide/Managing%20Volumes/#expanding-volumes
>
> Regards,
> Karthik
>
> On Fri, Apr 6, 2018 at 11:39 PM, Jose Sanchez <josesanc at carc.unm.edu>
> wrote:
>
>> Hi Karthik
>>
>> this is our configuration, is 2x2 =4 , they are all replicated , each
>> brick has 14tb. we have 2 nodes A and B, each one with brick 1 and 2.
>>
>> Node A (replicated A1 (14tb) and B1 (14tb) ) same with node B
>> (Replicated A2 (14tb) and B2 (14tb)).
>>
>> Do you think we need to degrade the node first before removing it. i
>> believe the same copy of data is on all 4 bricks, we would like to keep one
>> of them, and add the other bricks as extra space
>>
>> Thanks for your help on this
>>
>> Jose
>>
>>
>>
>>
>>
>> [root at gluster01 ~]# gluster volume info scratch
>>
>> Volume Name: scratch
>> Type: Distributed-Replicate
>> Volume ID: 23f1e4b1-b8e0-46c3-874a-58b4728ea106
>> Status: Started
>> Snapshot Count: 0
>> Number of Bricks: 2 x 2 = 4
>> Transport-type: tcp,rdma
>> Bricks:
>> Brick1: gluster01ib:/gdata/brick1/scratch
>> Brick2: gluster02ib:/gdata/brick1/scratch
>> Brick3: gluster01ib:/gdata/brick2/scratch
>> Brick4: gluster02ib:/gdata/brick2/scratch
>> Options Reconfigured:
>> performance.readdir-ahead: on
>> nfs.disable: on
>>
>> [root at gluster01 ~]# gluster volume status all
>> Status of volume: scratch
>> Gluster process TCP Port RDMA Port Online
>> Pid
>> ------------------------------------------------------------
>> ------------------
>> Brick gluster01ib:/gdata/brick1/scratch 49152 49153 Y
>> 1743
>> Brick gluster02ib:/gdata/brick1/scratch 49156 49157 Y
>> 1732
>> Brick gluster01ib:/gdata/brick2/scratch 49154 49155 Y
>> 1738
>> Brick gluster02ib:/gdata/brick2/scratch 49158 49159 Y
>> 1733
>> Self-heal Daemon on localhost N/A N/A Y
>> 1728
>> Self-heal Daemon on gluster02ib N/A N/A Y
>> 1726
>>
>> Task Status of Volume scratch
>> ------------------------------------------------------------
>> ------------------
>> There are no active volume tasks
>>
>> ---------------------------------
>> Jose Sanchez
>> Systems/Network Analyst 1
>> Center of Advanced Research Computing
>> 1601 Central Ave
>> <https://maps.google.com/?q=1601+Central+Ave&entry=gmail&source=g>.
>> MSC 01 1190
>> Albuquerque, NM 87131-0001
>> carc.unm.edu
>> 575.636.4232
>>
>> On Apr 6, 2018, at 3:49 AM, Karthik Subrahmanya <ksubrahm at redhat.com>
>> wrote:
>>
>> Hi Jose,
>>
>> By switching into pure distribute volume you will lose availability if
>> something goes bad.
>>
>> I am guessing you have a nX2 volume.
>> If you want to preserve one copy of the data in all the distributes, you
>> can do that by decreasing the replica count in the remove-brick operation.
>> If you have any inconsistency, heal them first using the "gluster volume
>> heal <volname>" command and wait till the
>> "gluster volume heal <volname> info" output becomes zero, before removing
>> the bricks, so that you will have the correct data.
>> If you do not want to preserve the data then you can directly remove the
>> bricks.
>> Even after removing the bricks the data will be present in the backend of
>> the removed bricks. You have to manually erase them (both data and
>> .glusterfs folder).
>> See [1] for more details on remove-brick.
>>
>> [1]. https://docs.gluster.org/en/latest/Administrator%20Guid
>> e/Managing%20Volumes/#shrinking-volumes
>>
>> HTH,
>> Karthik
>>
>>
>> On Thu, Apr 5, 2018 at 8:17 PM, Jose Sanchez <josesanc at carc.unm.edu>
>> wrote:
>>
>>>
>>> We have a Gluster setup with 2 nodes (distributed replication) and we
>>> would like to switch it to the distributed mode. I know the data is
>>> duplicated between those nodes, what is the proper way of switching it to a
>>> distributed, we would like to double or gain the storage space on our
>>> gluster storage node. what happens with the data, do i need to erase one of
>>> the nodes?
>>>
>>> Jose
>>>
>>>
>>> ---------------------------------
>>> Jose Sanchez
>>> Systems/Network Analyst
>>> Center of Advanced Research Computing
>>> 1601 Central Ave
>>> <https://maps.google.com/?q=1601+Central+Ave&entry=gmail&source=g>.
>>> MSC 01 1190
>>> Albuquerque, NM 87131-0001
>>> carc.unm.edu
>>> 575.636.4232
>>>
>>>
>>> _______________________________________________
>>> Gluster-users mailing list
>>> Gluster-users at gluster.org
>>> http://lists.gluster.org/mailman/listinfo/gluster-users
>>>
>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20180412/60248984/attachment.html>
More information about the Gluster-users
mailing list