[Lustre-discuss] LustreError: The mds_getxattr operation failed with -43

Gonçalo Borges goncalo at lip.pt
Mon Jan 12 10:21:35 PST 2009


Hi All...

I'm having the following problems:

- It seems my clients are not able to reach my mdt. If you do a "dmesg" 
in a client linux machine, you will get:

---*---
LustreError: 11-0: an error occurred while communicating with 
172.30.1.209 at tcp. The mds_getxattr operation failed with -43
LustreError: Skipped 1 previous similar message
LustreError: 2472:0:(dir.c:388:ll_readdir()) error reading dir 
6422529/2008980280 page 0: rc -43
LustreError: 2623:0:(dir.c:388:ll_readdir()) error reading dir 
6422529/2008980280 page 0: rc -43
---*---

Probably, as a consequence, sometimes some of the users can not list 
directory contents. Check the following test:

[cmsprd081 at srm01 ~]$ ll 
/lustre/lip.pt/data/cms/store/unmerged/SAM/testSRM/SAM-srm01.lip.pt/lcg-util 

ls: 
/lustre/lip.pt/data/cms/store/unmerged/SAM/testSRM/SAM-srm01.lip.pt/lcg-util: 
Identifier removed
[cmsprd081 at srm01 ~]$ exit
logout

[root at srm01 ~]# ll 
/lustre/lip.pt/data/cms/store/unmerged/SAM/testSRM/SAM-srm01.lip.pt/lcg-util 

total 0
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 testfile-cp-20090112-170527.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 testfile-cp-20090112-180358.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 
testfile-cp-CMS_DEFAULT-20090112-170527.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-cp-CMS_DEFAULT-20090112-180358.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 testfile-gt-20090112-170618.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 testfile-gt-20090112-180445.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-gt-CMS_DEFAULT-20090112-170618.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-CMS_DEFAULT-20090112-180445.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-gt-rm-gt-20090112-170557.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-rm-gt-20090112-180427.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 
testfile-gt-rm-gt-CMS_DEFAULT-20090112-170557.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-rm-gt-CMS_DEFAULT-20090112-180427.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 testfile-ls-20090112-170627.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 testfile-ls-20090112-180516.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-ls-CMS_DEFAULT-20090112-170627.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 
testfile-ls-CMS_DEFAULT-20090112-180516.txt

[root at srm01 ~]# su -l cmsprd081
[cmsprd081 at srm01 ~]$ ll 
/lustre/lip.pt/data/cms/store/unmerged/SAM/testSRM/SAM-srm01.lip.pt/lcg-util 

total 0
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 testfile-cp-20090112-170527.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 testfile-cp-20090112-180358.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 
testfile-cp-CMS_DEFAULT-20090112-170527.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-cp-CMS_DEFAULT-20090112-180358.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 testfile-gt-20090112-170618.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 testfile-gt-20090112-180445.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-gt-CMS_DEFAULT-20090112-170618.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-CMS_DEFAULT-20090112-180445.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-gt-rm-gt-20090112-170557.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-rm-gt-20090112-180427.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:05 
testfile-gt-rm-gt-CMS_DEFAULT-20090112-170557.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:04 
testfile-gt-rm-gt-CMS_DEFAULT-20090112-180427.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 testfile-ls-20090112-170627.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 testfile-ls-20090112-180516.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 16:06 
testfile-ls-CMS_DEFAULT-20090112-170627.txt
-rw-rwx---+ 1 storm storm 0 Jan 12 17:05 
testfile-ls-CMS_DEFAULT-20090112-180516.txt
[cmsprd081 at srm01 ~]$

In summary, the first time the cmsprd081 user tried to list the 
directory, he got a "Identifier removed" error. The second time, he was 
able to do it...

I need suggestions here, or things to try...

Thanks in Advance

Gonçalo



More information about the lustre-discuss mailing list