[lustre-discuss] lustre 2.8.0, getattr error rc = -2

[Patrick Farrell]

Christopher,

Andreas is suggesting a Lustre patch - There is no relevant knob to turn
to control it, it’s just currently a type of message that should always go
to the console (a warning or emergency message).  You probably *don’t*
want to suppress those, which is why there’s no easy way to do that, and
why he’s suggesting a patch to change this message to a different type.

- Patrick

On 4/12/16, 9:32 AM, "lustre-discuss on behalf of Christopher B Coffey"
<lustre-discuss-bounces at lists.lustre.org on behalf of
Chris.Coffey at nau.edu> wrote:

>Thank you Andreas.  That makes good sense.  I’m sorry, how would you go
>about quieting that message from the console?  Possibly with sysctl ?  Or
>maybe with lctl?
>
>Best,
>Chris
>―
>Christopher Coffey
>High-Performance Computing
>Northern Arizona University
>928-523-1167
>
>
>
>
>
>
>
>
>On 4/11/16, 3:31 PM, "Dilger, Andreas" <andreas.dilger at intel.com> wrote:
>
>>On 2016/04/11, 16:01, "Christopher B Coffey" <Chris.Coffey at nau.edu>
>>wrote:
>>
>>>Hi,
>>>
>>>I¹ve upgraded our lustre filesystem from 2.5.3 (server), and 2.6.0
>>>(client) to 2.8.0 all around.  Things look great.
>>
>>Good to hear.
>>
>>>  But I am getting these errors on the mds logs:
>>>
>>>Apr 11 11:04:02 mds1 kernel: LustreError:
>>>43984:0:(mdt_handler.c:893:mdt_getattr_internal()) blizzard-MDT0000:
>>>getattr error for [0x200040406:0xeba3:0x0]: rc = -2
>>>
>>>
>>>But,
>>>
>>>lfs fid2path /scratch 0x200040406:0xeba3:0x0
>>>fid2path: error on FID 0x200040406:0xeba3:0x0: No such file or directory
>>>
>>>
>>>So, the file is really not there.  But, why would the client be
>>>requesting a file that is not there?  Maybe broken hardlink?  Or what
>>>else could it be?  A timing issue where the file was deleted in the
>>>process of being retrieved?  Or does this error message indicate
>>>something else? Thanks!
>>
>>Typically I've seen this when there are two processes running "rm -r" in
>>the same directory tree, or equivalent (at least one doing "rm", the
>>other
>>one could be anything like "find" or "ls -lR" or "rsync" ...).
>>
>>This particular message could be quieted from the console
>>(CERROR->CDEBUG), since it can be generated by a regular user easily and
>>does not necessarily indicate anything is wrong.  Please file a ticket in
>>Jira, and if you can also submit a patch that would be even better.
>>
>>Cheers, Andreas
>>-- 
>>Andreas Dilger
>>
>>Lustre Principal Architect
>>Intel High Performance Data Division
>>
>>
>_______________________________________________
>lustre-discuss mailing list
>lustre-discuss at lists.lustre.org
>http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org