Hi Nick, I forgot to mention that I was also trying a workaround using the userland (rbd-fuse). The behaviour was exactly the same (worked fine for several hours, testing parallel reading and writing, then IO Wait and system load increased). This is why I don't think it is an issue with the rbd kernel module. Regards, Christian Am 20.04.2015 um 11:37 schrieb Nick Fisk: > Hi Christian, > > A very non-technical answer but as the problem seems related to the RBD > client it might be worth trying the latest Kernel if possible. The RBD > client is Kernel based and so there may be a fix which might stop this from > happening. > > Nick > >> -----Original Message----- >> From: ceph-users [mailto:ceph-users-bounces@xxxxxxxxxxxxxx] On Behalf Of >> Christian Eichelmann >> Sent: 20 April 2015 08:29 >> To: ceph-users@xxxxxxxxxxxxxx >> Subject: 100% IO Wait with CEPH RBD and RSYNC >> >> Hi Ceph-Users! >> >> We currently have a problem where I am not sure if the it has it's cause > in >> Ceph or something else. First, some information about our ceph-setup: >> >> * ceph version 0.87.1 >> * 5 MON >> * 12 OSD with 60x2TB each >> * 2 RSYNC Gateways with 2x10G Ethernet (Kernel: 3.16.3-2~bpo70+1, Debian >> Wheezy) >> >> Our cluster is mainly used to store Log-Files from numerous servers via > RSync >> and make them available via RSync as well. Since about two weeks we have a >> very strange behaviour and our RSync Gateways (they just map several rbd >> devices and "export" them via rsyncd): The IO Wait on the systems are >> increasing untill some of the cores getting stuck with an IO Wait of 100%. >> RSync processes become zombies (defunct) and/or can not be killed even >> with SIGKILL. After the system has reached a load of about 1400, it > becomes >> totally unresponsive and the only way to "fix" the problem is to reboot > the >> system. >> >> I was trying to manually reproduce the problem by simultainously reading >> and writing from several machine, but the problem didn't appear. >> >> I have no idea where the error can be. I was doing a ceph tell osd.* bench >> during the problem and all osds where having normal benchmark results. Has >> anyone an idea how this can happen? If you need any more informations, >> please let me know. >> >> Regards, >> Christian >> >> _______________________________________________ >> ceph-users mailing list >> ceph-users@xxxxxxxxxxxxxx >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > > > -- Christian Eichelmann Systemadministrator 1&1 Internet AG - IT Operations Mail & Media Advertising & Targeting Brauerstraße 48 · DE-76135 Karlsruhe Telefon: +49 721 91374-8026 christian.eichelmann@xxxxxxxx Amtsgericht Montabaur / HRB 6484 Vorstände: Henning Ahlert, Ralph Dommermuth, Matthias Ehrlich, Robert Hoffmann, Markus Huhn, Hans-Henning Kettler, Dr. Oliver Mauss, Jan Oetjen Aufsichtsratsvorsitzender: Michael Scheeren _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |