[Gluster-devel] a question again , about the dht

Shylesh Kumar shmohan at redhat.com
Thu Jan 23 06:10:29 UTC 2014


Krishnan Parthasarathi wrote:
> Xinguo,
>
> dht translator is responsible for the "trusted.glusterfs.dht.*" extended
> attributes. The code for this is present at xlators/cluster/dht/src relative
> to the root of the code repo.
>
> Did you perform any rebalance or remove-brick operation after starting
> the brick and before writing any data on the volume via a mount?
> These are other processes that include dht translator in their runtime.
I suspect gluster-nfs process does also load the dht which will be started
at the time of volume start.
>
> thanks,
> Krish
>
>
>
> ----- Original Message -----
>> Krish,
>>
>> Thanks for your help ,but my circumstance is that the "tursted.glusterfs.dht"
>>   xattr is set just without the mount process running. I know that the mount
>> process can do the hash-range-assign job , but in my case , the mount
>> process at the client side is not being run  when I do the "volume start"
>> commond at the server side , and the "tursted.glusterfs.dht" xattr can be
>> set correctly  too. So I guess that there must be some other process at the
>> server side did the hash-range-assign job, I just want to find out which
>> process did it. Can you give me some information about this, thanks!
>>
>> Thanks again,
>> Xinguo
>>
>> At 2014-01-21 01:18:37,"Krishnan Parthasarathi" <kparthas at redhat.com> wrote:
>>> Xinguo,
>>>
>>> You should be attaching gdb to the mount process. The process serves
>>>   filesystem requests on the mount point. This process has the mount point
>>>   (directory) in its command line arguments. This should help you find out
>>>   the pid of the process using ps(1).
>>>
>>> Hope that helps,
>>> krish
>>>
>>> ----- Original Message -----
>>>>   Hello,
>>>>   
>>>>   In glusterfs server side , when we do the "volume start" commond , every
>>>>   brick's root directory will be set the "tursted.glusterfs.dht" xattr and
>>>>   will be assigned a hash-range , as soon as the commond returned
>>>>    successfully
>>>>   . we know that the function "dht_selfheal_layout_new_directory()" do the
>>>>   hash-range assign work , but when I make a breakpoint at function
>>>>   "dht_selfheal_layout_new_directory()" in the attached process "glusterd"
>>>>   before doing the "volume start" commond , the "glusterd" process won't
>>>>    stop
>>>>   at the breakpoint and the "tursted.glusterfs.dht" xattr of every brick's
>>>>   root directory still be set successfully . Why ? Did I attached the wrong
>>>>   process ? What shoud I do to let the hash-range assigning work stop so I
>>>>    can
>>>>   follow the assigning work step by step ?
>>>>   
>>>>   hope to get help,
>>>>   thanks,
>>>>   Xinguo
>>>>   
>>>>   
>>>>   
>>>>   _______________________________________________
>>>>   Gluster-devel mailing list
>>>>   Gluster-devel at nongnu.org
>>>>   https://lists.nongnu.org/mailman/listinfo/gluster-devel
>>>>
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> https://lists.nongnu.org/mailman/listinfo/gluster-devel





More information about the Gluster-devel mailing list