Oh and I wasn't say we shouldnt include it, we just need to be aware
of the limitations here and note that these kind of 'ssh back'
solutions has problems at the 15-20 node scale.  (having done this
before I have experienced it...)

the jira is just awaiting a patch for someone who wants to be famous :-)

On Mon, Sep 20, 2010 at 10:17 PM, Ryan Rawson <[email protected]> wrote:
> I dont have a solution for an extremely large cluster, serial rsync
> works for 20-40+ nodes, I dont have any solutions beyond that.
>
> -ryan
>
> On Mon, Sep 20, 2010 at 10:10 PM, Jeff Zhang <[email protected]> wrote:
>> Hi Ryan,
>>
>> Then what is the solution for large cluster like you said having 20+
>> machines ? Thanks, and I curious to know how yahoo solve this for his
>> extremely large cluster ?
>>
>>
>>
>> On Tue, Sep 21, 2010 at 12:27 PM, Ryan Rawson <[email protected]> wrote:
>>> This kind of script is not general, so one would want to make sure it
>>> is off by default.  This is because having 20+ machines ssh back to a
>>> singular master triggers the sshd dos protection :-)
>>>
>>> -ryan
>>>
>>> On Mon, Sep 20, 2010 at 8:46 PM, Ted Yu <[email protected]> wrote:
>>>> I raised the same question 4 months ago:
>>>> https://issues.apache.org/jira/browse/HBASE-2789
>>>>
>>>> On Mon, Sep 20, 2010 at 6:24 PM, Jeff Zhang <[email protected]> wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> I am not sure why hbase do not have rsync in start-hbase.sh ? In my
>>>>> opinion, this feature is very useful for users, is there any other
>>>>> consideration about this  ?
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Best Regards
>>>>>
>>>>> Jeff Zhang
>>>>>
>>>>
>>>
>>
>>
>>
>> --
>> Best Regards
>>
>> Jeff Zhang
>>
>

Reply via email to