[Gluster-users] geo replication, invalid slave name and gluster 3.5.1

Tue Jul 15 15:15:23 UTC 2014

On 15/07/14 18:13, Stefan Moravcik wrote:
> Hello Vishwanath
>
> thank you for your quick reply but i have a follow up question if it 
> is ok... Maybe a different issue and i should open a new thread, but i 
> will try to continue to use this one...
>
> So I followed the new documentation... let me show you what i have 
> done and what is the final error message...
>
>
> I have 3 servers node1, node2 and node3 with IPs 1.1.1.1, 1.1.1.2 and 
> 1.1.1.3
>
> I installed glusterfs-server and glusterfs-geo-replication on all 3 of 
> them... I created replica volume called myvol1 and run the command
>
> gluster system:: execute gsec_create
>
> this created 4 files:
> secret.pem
> secret.pem.pub
> tar_ssh.pem
> tar_ssh.pem.pub
>
> The pub file is different on all 3 nodes so I copied all 3 
> secret.pem.pub to slave authorized_keys. I tried to ssh directly to 
> slave server from all 3 nodes and got through with no problem.
>
> So I connected to slave server installed glusterfs-server and 
> glusterfs-geo-replication there too.
>
> Started the glusterd and created a volume called myvol1_slave
>
> Then I peer probed one of the masters with slave. This showed the 
> volume in my master and peer appeared in peer status.
>
> From here i run the command in your documentation
>
> volume geo-replication myvol1 1.2.3.4::myvol1_slave create push-pem
> Passwordless ssh login has not been setup with 1.2.3.4.
> geo-replication command failed
Couple of things here.

I believe it was not clear enough in the docs and I apologise for that. 
But this is the prerequisite for dist-geo-rep.

* /There should be a password-less ssh setup between at least one node 
in master volume to one node in slave volume. The geo-rep create command 
should be executed from this node which has password-less ssh setup to 
slave./

So in your case, you can setup a password less ssh between 1.1.1.1 (one 
master volume node) to 1.2.3.4 (one slave volume node). You can use 
"ssh-keygen" and "ssh-copy-id" to do the same.
After the above step is done, execute the "gluster system:: execute 
gsec_create". You don't need to copy it to the slave autorized_keys. 
geo-rep create push-pem takes care of it for you.

Now, you should execute "gluster volume geo-rep myvol1 
1.2.3.4::myvol1_slave cerate push-pem" from 1.1.1.1 (because this node 
has passwordless ssh to 1.2.3.4 mentioned in the command)

That should create a geo-rep session for you. That can be started later on.

And you don't need to peer probe slave from master or vice versa. 
Logically both master and slave volumes are in different clusters (in 
two different geographic locations).

HTH,
Vishwanath

>
> In the secure log file i could see the connection though.
>
> 2014-07-15T13:26:56.083445+01:00 1testlab sshd[23905]: Set 
> /proc/self/oom_score_adj to 0
> 2014-07-15T13:26:56.089423+01:00 1testlab sshd[23905]: Connection from 
> 1.1.1.1 port 58351
> 2014-07-15T13:26:56.248687+01:00 1testlab sshd[23906]: Connection 
> closed by 1.1.1.1
>
> and in the logs of one of the masters
>
> [2014-07-15 12:26:56.247667] E 
> [glusterd-geo-rep.c:1889:glusterd_verify_slave] 0-: Not a valid slave
> [2014-07-15 12:26:56.247752] E 
> [glusterd-geo-rep.c:2106:glusterd_op_stage_gsync_create] 0-: 
> 1.2.3.4::myvol1_slave is not a valid slave volume. Error: Passwordless 
> ssh login has not been setup with 1.2.3.4.
> [2014-07-15 12:26:56.247772] E 
> [glusterd-syncop.c:912:gd_stage_op_phase] 0-management: Staging of 
> operation 'Volume Geo-replication Create' failed on localhost : 
> Passwordless ssh login has not been setup with 1.2.3.4.
>
> there is no log in the other masters in the cluster nor on slave..
>
> I even tried with force option, but same result... I disabled firewall 
> and selinux just to make sure those parts of the system do not 
> interfere. Searched a google for same problem and found one... 
> http://irclog.perlgeek.de/gluster/2014-01-16 but again no answer or 
> solution.
>
> Thank you for your time and help.
>
> Best regards,
> Stefan
>
> On 15/07/14 12:26, M S Vishwanath Bhat wrote:
>> On 15/07/14 15:08, Stefan Moravcik wrote:
>>> Hello Guys,
>>>
>>> I have been trying to set a geo replication in our glusterfs test 
>>> environment and got a problem with a message "invalid slave name"
>>>
>>> So first things first...
>>>
>>> I have 3 nodes configured in a cluster. Those nodes are configured 
>>> as replica. On this cluster I have a volume created with let say 
>>> name myvol1. So far everything works and looks good...
>>>
>>> Next step was to create a geo replication off site.. So i followed 
>>> this documentation:
>>> http://www.gluster.org/community/documentation/index.php/HowTo:geo-replication 
>>>
>> These are old docs. I have edited this to mention that it is old 
>> geo-rep docs.
>>
>> Please refer to 
>> https://github.com/gluster/glusterfs/blob/master/doc/admin-guide/en-US/markdown/admin_distributed_geo_rep.md 
>> or 
>> https://medium.com/@msvbhat/distributed-geo-replication-in-glusterfs-ec95f4393c50 
>> for latest distributed-geo-rep documentation.
>>>
>>> I had peered the slave server, created secret.pem was able to ssh 
>>> without the password and tried to create the geo replication volume 
>>> with the code from the documentation and got the following error:
>>>
>>> on master:
>>> gluster volume geo-replication myvol1 1.2.3.4:/shared/myvol1_slave 
>>> start
>>>
>>> on master:
>>> [2014-07-15 09:15:37.188701] E 
>>> [glusterd-geo-rep.c:4083:glusterd_get_slave_info] 0-: Invalid slave 
>>> name
>>> [2014-07-15 09:15:37.188827] W [dict.c:778:str_to_data] 
>>> (-->/usr/lib64/glusterfs/3.5.1/xlator/mgmt/glusterd.so(glusterd_op_stage_gsync_create+0x1e2) 
>>> [0x7f979e20f1f2] 
>>> (-->/usr/lib64/glusterfs/3.5.1/xlator/mgmt/glusterd.so(glusterd_get_slave_details_confpath+0x116) 
>>> [0x7f979e20a306] (-->/usr/lib64/libglusterfs.so.0(dict_set_str+0x1c) 
>>> [0x7f97a322045c]))) 0-dict: value is NULL
>>> [2014-07-15 09:15:37.188837] E 
>>> [glusterd-geo-rep.c:3995:glusterd_get_slave_details_confpath] 0-: 
>>> Unable to store slave volume name.
>>> [2014-07-15 09:15:37.188849] E 
>>> [glusterd-geo-rep.c:2056:glusterd_op_stage_gsync_create] 0-: Unable 
>>> to fetch slave or confpath details.
>>> [2014-07-15 09:15:37.188861] E 
>>> [glusterd-syncop.c:912:gd_stage_op_phase] 0-management: Staging of 
>>> operation 'Volume Geo-replication Create' failed on localhost
>>>
>>> there are no logs on slave what so ever
>>> I also tried different documentation with "create push-pem" got the 
>>> very same problem as above...
>>>
>>> I tried to start the volume as node:/path/to/dir and also created a 
>>> volume on slave and started as node:/slave_volume_name always a same 
>>> result...
>>>
>>> Tried to search for a solution and found this 
>>> http://fpaste.org/114290/04117421/
>>>
>>> It was different user with a very same problem... The issue was 
>>> shown on IRC channel, but never answered..
>>>
>>> This is a fresh install of 3.5.1, so no upgrade should be needed i 
>>> guess... Any help solving this problem would be appreciated..
>> From what you have described, it looks like your slave is not a 
>> gluster volume. In latest geo-rep, slave has to be a gluster volume. 
>> Now glusterfs does not support a simple directory as a slave.
>>
>> Please follow new documentation and try once more.
>>
>> HTH
>>
>> Best Regards,
>> Vishwanath
>>
>>>
>>> Thank you and best regards,
>>> Stefan
>>>
>>>
>>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140715/09614613/attachment.html>