Wilko just told me that this won't work --- because your data server is still in the xrootd cluster and redirector will send you back to the data server that you want to drain. A slightly modify way will work:
xrdcp file root://redirector//file.TMP
use xrd to rename root://redirector//file.TMP to root://redirector//file
regards,
Wei Yang | [log in to unmask] | 650-926-3338(O)
On Jan 12, 2012, at 12:15 PM, Yang, Wei wrote:
> Hi Nate,
>
> You can use cms.space directive set the minimum required free space to, say 1000PB and restart cmsd/xrootd. Your redirector will then think that this data server run out of space and will not send future writing to it. After than you can drain your data server.
>
> Another way to do this is to set r/o in all.export
>
> regards,
> Wei Yang | [log in to unmask] | 650-926-3338(O)
>
>
> On Jan 12, 2012, at 11:46 AM, Nathan Yehle wrote:
>
>> Wei,
>>
>> Is there a way to drain an xrootd server without bringing it offline? (stopping xrootd/cmsd on a production server is not an option unfortunately)
>>
>> -Nate
>>
>> On Jan 9, 2012, at 3:04 AM, Yang, Wei wrote:
>>
>>> Here is another way to do this, on the "faulty" data sever (obviously, it can't be complete faulty, otherwise you can't login and do anything)
>>>
>>> stop xrootd/cmsd
>>> find storage_root_path -type f | while read file; do
>>> dstfile=`remove local_root from $file`
>>> xrdcp $file root://your_redirector/$dstfile
>>> done
>>>
>>> you can speed up the process by making a more sophisticated script using xprep, etc.
>>>
>>> regards,
>>> Wei Yang | [log in to unmask] | 650-926-3338(O)
>>>
>>>
>>>
>>>
>>> On Jan 9, 2012, at 12:41 AM, Andrew Hanushevsky wrote:
>>>
>>>> Hi Doug,
>>>>
>>>> The short answer is no. On the positive side, we will submit an ASCR
>>>> proposal to allow that.
>>>>
>>>> Andy
>>>>
>>>> P.S. The current model is that missing files from a faulty data server
>>>> would be automatically copied in on other servers when so requested. That
>>>> is a reasonable model is you assume that te ailure rate is relatively
>>>> small.
>>>>
>>>> On Sat, 7 Jan 2012, Doug Benjamin wrote:
>>>>
>>>>> Hello,
>>>>>
>>>>> Are there any xrootd commands to drain a faulty data server of files
>>>>> and distribute them across the
>>>>> rest of the data servers in the storage system.
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Doug Benjamin
>>>>>
>>>>> ########################################################################
>>>>> Use REPLY-ALL to reply to list
>>>>>
>>>>> To unsubscribe from the XROOTD-L list, click the following link:
>>>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>>>>>
>>>>
>>>> ########################################################################
>>>> Use REPLY-ALL to reply to list
>>>>
>>>> To unsubscribe from the XROOTD-L list, click the following link:
>>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>>>
>>> ########################################################################
>>> Use REPLY-ALL to reply to list
>>>
>>> To unsubscribe from the XROOTD-L list, click the following link:
>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>>
>
> ########################################################################
> Use REPLY-ALL to reply to list
>
> To unsubscribe from the XROOTD-L list, click the following link:
> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the XROOTD-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
|