That kernel probably has the bug where tcp_nodelay is not enabled. That is fixed in Kernel 4.0+, however also in 4.0 blk-mq was introduced which brings two other limitations:-
1. Max queue depth of 128 2. IO’s sizes are restricted/split to 128kb From: ceph-users [mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of German Anders Sent: 24 June 2015 15:45 To: ceph-users Subject: [ceph-users] kernel 3.18 io bottlenecks? Hi all, Is there any IO botleneck reported on kernel 3.18.3-031803-generic? since I'm having a lot of iowait and the cluster is really getting slow, and actually there's no much going on. I've read some time ago that there were some issues with kern 3.18, so I would like to know what's the 'best' kernel to go with, I'm using Ubuntu 14.04.1LTs and Ceph v0.82. Thanks a lot, Best regards, German
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
