Hi,
we have observed that in some situations ( probably network glitches )
the pnfs client blacklisted one of the data servers:
NFS: data server 83a95099 connection error -12. Deviceid [22000000000]
marked out of use.
As a result, data server can't be used by this client anymore.
Is there a way to let client to forget about data server?
Some magic in /proc ?
This is SL6.2 (RHEL 6.2):
# uname -a
Linux p3-wgs13 2.6.32-220.2.1.el6.x86_64 #1 SMP Thu Dec 22 11:15:52
CST 2011 x86_64 x86_64 x86_64 GNU/Linux
#
Regards,
Tigran.
On 01/25/2012 02:44 PM, Boaz Harrosh wrote:
> On 01/25/2012 11:56 AM, Tigran Mkrtchyan wrote:
>> Hi,
>>
>> we have observed that in some situations ( probably network glitches )
>> the pnfs client blacklisted one of the data servers:
>>
>> NFS: data server 83a95099 connection error -12. Deviceid [22000000000]
>> marked out of use.
>>
>> As a result, data server can't be used by this client anymore.
>>
>> Is there a way to let client to forget about data server?
>> Some magic in /proc ?
>>
>> This is SL6.2 (RHEL 6.2):
>> # uname -a
>> Linux p3-wgs13 2.6.32-220.2.1.el6.x86_64 #1 SMP Thu Dec 22 11:15:52
>> CST 2011 x86_64 x86_64 x86_64 GNU/Linux
>> #
>>
>
> Look in the source code, I think there is a RECALL that the server
> can do to trash the all device cache. or one of the devices.
>
> What happens is that the device is marked with error but is in
> cache so is not re-fetched.
>
> wait let me look ....
>
> I found it! The server sends a NOTIFY_DEVICEID4_CHANGE. The
> client will remove the deviceid from cache and unmount if needed.
> Next layout with that deviceid will re-establish the connection and
> will put a new clean entry in the dev cache.
>
If you want to see for your self look at:
callback_proc.c::nfs4_callback_devicenotify()
Boaz
> [If you decide to enhance pynfs to send a NOTIFY_DEVICEID4_CHANGE as an admin
> tool. That would be interesting]
>
>> Regards,
>> Tigran.
>
> Cheers
> Boaz
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
On 01/25/2012 11:56 AM, Tigran Mkrtchyan wrote:
> Hi,
>
> we have observed that in some situations ( probably network glitches )
> the pnfs client blacklisted one of the data servers:
>
> NFS: data server 83a95099 connection error -12. Deviceid [22000000000]
> marked out of use.
>
> As a result, data server can't be used by this client anymore.
>
> Is there a way to let client to forget about data server?
> Some magic in /proc ?
>
> This is SL6.2 (RHEL 6.2):
> # uname -a
> Linux p3-wgs13 2.6.32-220.2.1.el6.x86_64 #1 SMP Thu Dec 22 11:15:52
> CST 2011 x86_64 x86_64 x86_64 GNU/Linux
> #
>
Look in the source code, I think there is a RECALL that the server
can do to trash the all device cache. or one of the devices.
What happens is that the device is marked with error but is in
cache so is not re-fetched.
wait let me look ....
I found it! The server sends a NOTIFY_DEVICEID4_CHANGE. The
client will remove the deviceid from cache and unmount if needed.
Next layout with that deviceid will re-establish the connection and
will put a new clean entry in the dev cache.
[If you decide to enhance pynfs to send a NOTIFY_DEVICEID4_CHANGE as an admin
tool. That would be interesting]
> Regards,
> Tigran.
Cheers
Boaz
Thanks Boaz I will check.
I believe rhel6 kernel does not support device notification.
Currently we just generated a new device id.
Tigran.
On Wed, Jan 25, 2012 at 1:46 PM, Boaz Harrosh <[email protected]> wrote:
> On 01/25/2012 02:44 PM, Boaz Harrosh wrote:
>> On 01/25/2012 11:56 AM, Tigran Mkrtchyan wrote:
>>> Hi,
>>>
>>> we have observed that in some situations ( probably network glitches )
>>> the pnfs client blacklisted one of the data servers:
>>>
>>> NFS: data server 83a95099 connection error -12. Deviceid [22000000000]
>>> marked out of use.
>>>
>>> As a result, data server can't be used by this client anymore.
>>>
>>> Is there a way to let client to forget about data server?
>>> Some magic in /proc ?
>>>
>>> This is SL6.2 (RHEL 6.2):
>>> # uname -a
>>> Linux p3-wgs13 2.6.32-220.2.1.el6.x86_64 #1 SMP Thu Dec 22 11:15:52
>>> CST 2011 x86_64 x86_64 x86_64 GNU/Linux
>>> #
>>>
>>
>> Look in the source code, I think there is a RECALL that the server
>> can do to trash the all device cache. or one of the devices.
>>
>> What happens is that the device is marked with error but is in
>> cache so is not re-fetched.
>>
>> wait let me look ....
>>
>> I found it! The server sends a NOTIFY_DEVICEID4_CHANGE. The
>> client will remove the deviceid from cache and unmount if needed.
>> Next layout with that deviceid will re-establish the connection and
>> will put a new clean entry in the dev cache.
>>
>
> If you want to see for your self look at:
> callback_proc.c::nfs4_callback_devicenotify()
>
> Boaz
>> [If you decide to enhance pynfs to send a NOTIFY_DEVICEID4_CHANGE as an admin
>> tool. That would be interesting]
>>
>>> Regards,
>>> Tigran.
>>
>> Cheers
>> Boaz
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
On 01/25/2012 03:17 PM, Tigran Mkrtchyan wrote:
> Thanks Boaz I will check.
>
> I believe rhel6 kernel does not support device notification.
> Currently we just generated a new device id.
>
OK That code is pretty old. How much of the pnfs was ported?
What is the pnfs Kernel version the port was based on?
new-ids is smart too.
Thanks
Boaz
> Tigran.
>
> On Wed, Jan 25, 2012 at 1:46 PM, Boaz Harrosh <[email protected]> wrote:
>> On 01/25/2012 02:44 PM, Boaz Harrosh wrote:
>>> On 01/25/2012 11:56 AM, Tigran Mkrtchyan wrote:
>>>> Hi,
>>>>
>>>> we have observed that in some situations ( probably network glitches )
>>>> the pnfs client blacklisted one of the data servers:
>>>>
>>>> NFS: data server 83a95099 connection error -12. Deviceid [22000000000]
>>>> marked out of use.
>>>>
>>>> As a result, data server can't be used by this client anymore.
>>>>
>>>> Is there a way to let client to forget about data server?
>>>> Some magic in /proc ?
>>>>
>>>> This is SL6.2 (RHEL 6.2):
>>>> # uname -a
>>>> Linux p3-wgs13 2.6.32-220.2.1.el6.x86_64 #1 SMP Thu Dec 22 11:15:52
>>>> CST 2011 x86_64 x86_64 x86_64 GNU/Linux
>>>> #
>>>>
>>>
>>> Look in the source code, I think there is a RECALL that the server
>>> can do to trash the all device cache. or one of the devices.
>>>
>>> What happens is that the device is marked with error but is in
>>> cache so is not re-fetched.
>>>
>>> wait let me look ....
>>>
>>> I found it! The server sends a NOTIFY_DEVICEID4_CHANGE. The
>>> client will remove the deviceid from cache and unmount if needed.
>>> Next layout with that deviceid will re-establish the connection and
>>> will put a new clean entry in the dev cache.
>>>
>>
>> If you want to see for your self look at:
>> callback_proc.c::nfs4_callback_devicenotify()
>>
>> Boaz
>>> [If you decide to enhance pynfs to send a NOTIFY_DEVICEID4_CHANGE as an admin
>>> tool. That would be interesting]
>>>
>>>> Regards,
>>>> Tigran.
>>>
>>> Cheers
>>> Boaz
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html