[Gluster-devel] about afr

[Krishna Srinivas]

Nicolas,

It might be a bug. Let me try to reproduce the problem here and get back to you.

Krishna

On Wed, Jan 14, 2009 at 6:59 PM, nicolas prochazka
<prochazka.nicolas at gmail.com> wrote:
> hello again,
> To finish with this issue and information I can send you :
> If i stop glusterfsd  ( on server B) before to stop this server ( hard
> poweroff by pressed on/off ) , the problem does not occur.  If i hard
> poweroff without stop gluster ( a real crash ) problem occur .
> Regards
> Nicolas Prochazka.
>
> 2009/1/14 nicolas prochazka <prochazka.nicolas at gmail.com>
>>
>> hi again,
>> I continue my tests and :
>> In my case, if one file is open on gluster mount during stop of one afr
>> server,
>> gluster mount can not be acces ( gap ? ) in this server. All other client
>> ( C for example) which not opening file during stop, isn't affect, i can do
>> a ls or open after transport timeout time.
>> If i kill the process that's use this file, then i can using gluster mount
>> point without problem.
>>
>> Regards,
>> Nicolas Prochazka.
>>
>> 2009/1/12 nicolas prochazka <prochazka.nicolas at gmail.com>
>>>
>>> for your attention,
>>> it seems that's this problem occur only when files is open and use and
>>> gluster mount point .
>>> I use big files of computation ( ~ 10G)  with in the most important part,
>>> read. In this case problem occurs.
>>> If i using only small files which create only some time, no problem
>>> occur, gluster mount can use other afr server.
>>>
>>> Regards,
>>> Nicolas Prochazka
>>>
>>>
>>>
>>> 2009/1/12 nicolas prochazka <prochazka.nicolas at gmail.com>
>>>>
>>>> Hi,
>>>> I'm tryning to set
>>>> option transport-timeout 5
>>>> in protocol/client
>>>>
>>>> so a max of 10 seconds before restoring gluster in normal situation ?
>>>> no success, i always in the same situation, a 'ls /mnt/gluster'   not
>>>> respond after > 10 mins
>>>> I can not reuse glustermount exept kill glusterfs process.
>>>>
>>>> Regards
>>>> Nicolas Prochazka
>>>>
>>>>
>>>>
>>>> 2009/1/12 Raghavendra G <raghavendra at zresearch.com>
>>>>>
>>>>> Hi Nicolas,
>>>>>
>>>>> how much time did you wait before concluding the mount point to be not
>>>>> working? afr waits for a maximum of (2 * transport-timeout) seconds before
>>>>> returning sending reply to the application. Can you wait for some time and
>>>>> check out is this the issue you are facing?
>>>>>
>>>>> regards,
>>>>>
>>>>> On Mon, Jan 12, 2009 at 7:49 PM, nicolas prochazka
>>>>> <prochazka.nicolas at gmail.com> wrote:
>>>>>>
>>>>>> Hi.
>>>>>> I've installed this model to test Gluster :
>>>>>>
>>>>>> + 2 servers ( A B )
>>>>>>    - with glusterfsd  server  ( glusterfs--mainline--3.0--patch-842 )
>>>>>>    - with glusterfs  client
>>>>>> server conf file .
>>>>>>
>>>>>> + 1 server C only client mode.
>>>>>>
>>>>>> My issue :
>>>>>> If C open big file in this client configuration and then i stop server
>>>>>> A (or B )
>>>>>> gluster mount point on server C seems to be block, i can not do 'ls
>>>>>> -l'  for example.
>>>>>> Is a this thing is normal ? as C open his file on A or B , then it is
>>>>>> blocking when server down ?
>>>>>> I was thinking in client AFR, client can reopen file/block an other
>>>>>> server , i'm wrong ?
>>>>>> Should use HA translator ?
>>>>>>
>>>>>> Regards,
>>>>>> Nicolas Prochazka.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> volume brickless
>>>>>> type storage/posix
>>>>>> option directory /mnt/disks/export
>>>>>> end-volume
>>>>>>
>>>>>> volume brick
>>>>>> type features/posix-locks
>>>>>> option mandatory on          # enables mandatory locking on all files
>>>>>> subvolumes brickless
>>>>>> end-volume
>>>>>>
>>>>>> volume server
>>>>>> type protocol/server
>>>>>> subvolumes brick
>>>>>> option transport-type tcp
>>>>>> option auth.addr.brick.allow 10.98.98.*
>>>>>> end-volume
>>>>>> ---------------------------
>>>>>>
>>>>>> client config
>>>>>> volume brick_10.98.98.1
>>>>>> type protocol/client
>>>>>> option transport-type tcp/client
>>>>>> option remote-host 10.98.98.1
>>>>>> option remote-subvolume brick
>>>>>> end-volume
>>>>>>
>>>>>> volume brick_10.98.98.2
>>>>>> type protocol/client
>>>>>> option transport-type tcp/client
>>>>>> option remote-host 10.98.98.2
>>>>>> option remote-subvolume brick
>>>>>> end-volume
>>>>>>
>>>>>> volume last
>>>>>> type cluster/replicate
>>>>>> subvolumes brick_10.98.98.1 brick_10.98.98.2
>>>>>> end-volume
>>>>>>
>>>>>> volume iothreads
>>>>>> type performance/io-threads
>>>>>> option thread-count 2
>>>>>> option cache-size 32MB
>>>>>> subvolumes last
>>>>>> end-volume
>>>>>>
>>>>>> volume io-cache
>>>>>> type performance/io-cache
>>>>>> option cache-size 1024MB             # default is 32MB
>>>>>> option page-size  1MB              #128KB is default option
>>>>>> option force-revalidate-timeout 2  # default is 1
>>>>>> subvolumes iothreads
>>>>>> end-volume
>>>>>>
>>>>>> volume writebehind
>>>>>> type performance/write-behind
>>>>>> option aggregate-size 256KB # default is 0bytes
>>>>>> option window-size 3MB
>>>>>> option flush-behind on      # default is 'off'
>>>>>> subvolumes io-cache
>>>>>> end-volume
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Gluster-devel mailing list
>>>>>> Gluster-devel at nongnu.org
>>>>>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Raghavendra G
>>>>>
>>>>
>>>
>>
>
>
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>
>