[Gluster-devel] High availability

[Kevan Benson]

If all your AFR ad Unify translators are defined on the server side, you 
have to have some form of failover on the client, such as a round-robin 
IP address between the servers, or heartbeat for a shared IP, etc.

If you put unify on the client side, or both unify and AFR on the client 
side, nothing else should be required.  I would advise you to set the 
transport timeout to something smaller though,even if it's just for 
testing purposes and you set it higher again later.  I find 10 seconds 
works fine for me.

CityToo - Benjamin Loire wrote:
> OK => If I unplug ethernet eth0, /mnt/glusterfs is no responding until I plug again the ethernet eth0
> 
> I didn't configure transport timeout and if don't plug again ethernet eth0, /mnt/glusterfs is no responding ...What is the time-period before command complete successfully because I have waiting about 30 secondes !
> Have I to necessarily to configure transport timeout ?
> 
> When you say "But after that time-period command should complete successfully", is condition to plug again ethernet eth0 ?
> 
> If you want say that command should complete successfully after a time-period slightly greater than transport timeout without to plug again ethernet eth0 (server), have I to use necessarily translator unify with "rr" in first before translator AFR, or is translator AFR alone does automatically high availability if one sublvolume is down ? 
> 
> PS: I have only 2 nodes server for my test.
> Should I have 3 nodes server to have high availability if I use AFR with option replicate *:2 or should I have high availability too if I have 2 nodes server with AFR with option replicate *:2 if I unplug ethernet eth0 server ?
> 
> Benjamin LOIRE
> 
> Le Mardi 30 Octobre 2007 12:55, Raghavendra G a écrit :
>> Hi Benjamin,
>>
>> It may not be responsive for a period slightly greater than transport
>> timeout. But after that time-period command should complete successfully.
>>
>> regards,
>> On Oct 30, 2007 1:27 AM, CityToo - Benjamin Loire <bloire at citytoo.com>
>> wrote:
>>
>>> If I unplug ethernet on server A, /mnt/glusterfs is no responding.
>>> Yet I use unify rr + AFR between server A and server B
>>>
>>> Is it normal for a high availability storage system like glusterfs ?
>>>
>>> Benjamin LOIRE
>>>
>>>
>>> _______________________________________________
>>> Gluster-devel mailing list
>>> Gluster-devel at nongnu.org
>>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>>>
>>
>>
> 


-- 

-Kevan Benson
-A-1 Networks