On Wed, Feb 1, 2017 at 5:19 PM, Dustin Lundquist <[email protected]> wrote: > Hi All, > > I recently upgraded a small three node cluster to Kraken and noticed some > unusual traffic from the monitor nodes to each of the other nodes with a > destination TCP port of zero. I tracked this down to one of the ceph-mgr > threads with strace. > > root@vhost10:~# ps -eLf | grep ceph-mgr > ceph 17443 1 17443 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17444 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17445 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17446 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17447 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17448 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17449 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17450 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17451 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17452 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17453 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17454 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17455 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17456 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17457 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17458 0 17 17:00 ? 00:00:02 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > ceph 17443 1 17459 0 17 17:00 ? 00:00:00 /usr/bin/ceph-mgr > -f --cluster ceph --id a --setuser ceph --setgroup ceph > root 17755 17649 17755 0 1 17:16 pts/3 00:00:00 grep ceph-mgr > root@vhost10:~# strace -p 17446 -e connect > Process 17446 attached > connect(24, {sa_family=AF_INET6, sin6_port=htons(0), inet_pton(AF_INET6, > "fdd3:62f9:a63:1:ec4:7aff:fe74:17ae", &sin6_addr), sin6_flowinfo=0, > sin6_scope_id=0}, 28) = -1 EINPROGRESS (Operation now in progress) > connect(24, {sa_family=AF_INET6, sin6_port=htons(0), inet_pton(AF_INET6, > "fdd3:62f9:a63:1:ec4:7aff:fe74:17ae", &sin6_addr), sin6_flowinfo=0, > sin6_scope_id=0}, 28) = -1 ECONNREFUSED (Connection refused) > connect(24, {sa_family=AF_INET6, sin6_port=htons(0), inet_pton(AF_INET6, > "fdd3:62f9:a63:1:ec4:7aff:fe74:17ae", &sin6_addr), sin6_flowinfo=0, > sin6_scope_id=0}, 28) = -1 EINPROGRESS (Operation now in progress) > connect(24, {sa_family=AF_INET6, sin6_port=htons(0), inet_pton(AF_INET6, > "fdd3:62f9:a63:1:ec4:7aff:fe74:17ae", &sin6_addr), sin6_flowinfo=0, > sin6_scope_id=0}, 28) = -1 ECONNREFUSED (Connection refused) > > > This installation is using IPv6, so this could be a contributing factor.
Hmm, I certainly haven't tested on ipv6, but all the mon client code is common between daemons so I'm not sure why the mgr daemon would be acting differently to the others. Is the ceph-mgr daemon getting the same config as everyone else (i.e. are your mon addresses etc in [global] section)? > Additionally there is a Python error loading cherrypy, in the ceph-mgr log: You can ignore this, it just means the REST API won't be running. John > > 2017-02-01 17:00:32.480292 7f0e25697700 1 mgr init Loading python module > 'rest' > 2017-02-01 17:00:32.520284 7f0e25697700 -1 mgr load Module not found: 'rest' > 2017-02-01 17:00:32.520302 7f0e25697700 -1 mgr init Error loading module > 'rest': (2) No such file or directory > 2017-02-01 17:00:32.520316 7f0e25697700 -1 mgr init Traceback (most recent > call last): > > File "/usr/lib/ceph/mgr/rest/__init__.py", line 2, in <module> > from module import * # NOQA > > File "/usr/lib/ceph/mgr/rest/module.py", line 23, in <module> > import cherrypy > > ImportError: No module named cherrypy > > > Thanks, > > > Dustin Lundquist > > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
