Hi all, We have pretty annoying intermittent error, it happens once in a while and I can''t find why. Maybe someone has an idea on what it can be. The error on the client is: LustreError: 25893:0:(mdc_locks.c:665:mdc_enqueue()) ldlm_cli_ enqueue error: -13 LustreError: 25893:0:(file.c:3347:ll_inode_revalidate_fini()) failure -13 inode 55574529 The inode is always the same: debugfs: ncheck 55574529 Inode Pathname 55574529 //ROOT MDS doesn''t have any error. The error on the client is produced by nagios check and returns DISK CRITICAL - /FS is not accessible: Permission denied The nest check will be OK. The check is check_disk and it runs "stat (p->name, &stat_buf[0])" The server version is: 2.4.1 Client: 1.8.6 Thanks, Alex.
On Mon, 16 Dec 2013 14:07:16 +0100 Laifer, Roland (SCC) wrote:> FYI, we also see once in a while similar LustreError messages but with > different Lustre versions: 2.4.1 on clients and 2.1.3, 2.1.5 or 2.4.1 > on servers. The messages mainly appear on login nodes of our clusters. > The inode or FID in the error messages belongs to the root of the > file systems: > # lfs fid2path pfs2wor1 [0x1d080001:0x86b7421d:0x0] > /Thanks Roland. Do you have any assumptions about reasons of these errors? Thanks, Alex.
Am 16.12.2013 16:38, schrieb Alexander Oltu:> On Mon, 16 Dec 2013 14:07:16 +0100 > Laifer, Roland (SCC) wrote: > >> FYI, we also see once in a while similar LustreError messages but with >> different Lustre versions: 2.4.1 on clients and 2.1.3, 2.1.5 or 2.4.1 >> on servers. The messages mainly appear on login nodes of our clusters. >> The inode or FID in the error messages belongs to the root of the >> file systems: >> # lfs fid2path pfs2wor1 [0x1d080001:0x86b7421d:0x0] >> / > > Thanks Roland. Do you have any assumptions about reasons of these > errors?no. I have no assumptions about reasons of these errors. I had hoped that they will disappear after the upgrade to Lustre 2.4.1 on the server side. Since this did not happen I just opened a problem at our Lustre support company. Regards, Roland