Author: Jitendra Yadav, last modified by Sanjamala Nayeem on May 7, 2015
Original Publication Date: May 7, 2015
MapR 2.x, 3.x ,4.x
While doing any read/write operation on MapRFS, the clent needs to connect to all the MFS nodes where the desired name or data containers exist. This problem comes when if client is not able to perform read/write ops on a MFS node, In that case we might see below errors at client side.
139721763559168 rpc err Connection reset by peer(104) 28.122 to 10.10.10.11:5660, fid 13156.251.593598, upd 0
2015-04-20 20:37:35,3602 ERROR Client fs/client/fileclient/cc/client.cc:3468 Thread: 139721763559168 rpc err Connection reset by peer(104) 28.122 to 10.10.10.11:5660, fid 7344.1198.432180, upd 0
2015-04-20 20:38:56,5342 ERROR Client fs/client/fileclient/cc/client.cc:3468
The error "Connection reset by peer(104) 28.122 to 10.10.10.11:5660,, fid 7344.1198.432180, upd 0" means errno 104 ECONNRESET which is system call on a socket. So it means that the node 10.10.10.11:5660 is dead or not reachable.
The "upd 0" means this is an attempt to fetch data from the server (as opposed to writing data to it).
1. Need to check whether the node 10.10.10.11 is reachable or not, if not then troubleshoot OS/network part.
2. Check if mfs is running on node 10.10.10.11 or not. If not then restart the MFS service.
ps -ef | grep mfs
3. It might also worth to check if mfs got crashed last time. If recent core available under /opt/cores with prefix "mfs" then there was a mfs crash happened. In that situation please collect support dump, core trace and core file and open support ticket.
ls -l /opt/cores.