Apologies for replying to myself, but a thought just occurred to me
(likely my only one today)- did Birmingham at any point migrate their
dpm headnode or dpns database host? That might be the problem.
Furthermore if things still don't work, check the DPM_HOST environmental
variable for the root shell in which you're firing off these deletes. It
should point to epgse1(.ph.bham.ac.uk).
Cheers!
Matt
On 02/12/15 13:56, Matt Doidge wrote:
> Hi Matt,
>
> On 02/12/15 13:43, Matt Williams wrote:
>> I guess I'm missing something simple but:
>>
>> rfrm -rf /dpm/ph.bham.ac.uk/home/atlas/atlasproddisk
>>
>> gives
>>
>> send2nsd: NS009 - fatal configuration error: Host unknown:
>> dpnshome.ph.bham.ac.uk
>> /dpm/ph.bham.ac.uk/home/atlas/atlasproddisk : Host not known
>>
>
> This command is right - we had a similar looking but subtly different
> error hit us at Lancaster a lot. For us it the rfrm was stalling when it
> hit a file that was recorded as having replicas on a disk server that
> was no longer with us (which stops the rfrm -rf). In that case the fix
> was to delete that file replica metadata:
> dpns-rm -af /dpm/naughty/file
>
> This looks a bit different - my advice would be to try to delete further
> within the proddisk directory tree, i.e. dpns-ls
> /dpm/ph.bham.ac.uk/home/atlas/atlasproddisk and set the rfrm at one of
> the sub-directories listed (best to not test this out with rucio though!).
>
> In the end I resorted to a dodgey script in called in a loop which
> looped over the sub-directories that I had left, recorded the files that
> it failed on, then dpns-rm -af'd the files for the next run. A few days
> running in a screen session finally ended our deletion woes.
>
> Hope that helps!
>
> Cheers,
> Matt
>
>> and if I do:
>>
>> rfrm -rf epgse1:/dpm/ph.bham.ac.uk/home/atlas/atlasproddisk
>>
>> I get
>>
>> epgse1:/dpm/ph.bham.ac.uk/home/atlas/atlasproddisk : No such file
>> or directory
>>
>> epgse1 is our DPM head node and I'm running the commands as root on
>> that machine.
>>
>> Any ideas?
>>
>> Cheers,
>> Matt
>>
>> On 30 October 2015 at 10:59, Alessandra Forti
>> <[log in to unmask]> wrote:
>>> The shortest command is
>>>
>>> rfrm -rf /dpm/$(domainname)/home/atlas/atlasproddisk
>>>
>>> it will take several hours and then some more so send it in back
>>> ground with
>>> some redirection in case of error. Or if you are neurotic like me you
>>> can do
>>> a number of subdirs in parallel. When it finishes
>>>
>>> dpm-releasespace --space_token ATLASPRODDISK
>>>
>>> cheers
>>> alessandra
>>>
>>>
>>> On 30/10/2015 10:45, Ewan MacMahon wrote:
>>>>>
>>>>> -----Original Message-----
>>>>> From: Testbed Support for GridPP member institutes [mailto:TB-
>>>>> [log in to unmask]] On Behalf Of Sam Skipsey
>>>>>
>>>>> Just to double-check (given that we all seem to have between 2 and
>>>>> 10 TB
>>>>> of stuff in our PRODDISKs), it's okay to delete the remaining
>>>>> contents?
>>>>>
>>>> And for the benefit of the busy/lazy[1], what would be the
>>>> recommended way
>>>> of doing that on a DPM?
>>>>
>>>> Ewan
>>>>
>>>>
>>>> [1] i.e. me.
>>>
>>>
>>> --
>>> Respect is a rational process. \\//
>>> Fatti non foste a viver come bruti (Dante)
|