I''m curious if anybody else has seen this behavior.... Right now I have puppetd making sure that NFS and NTP services are running across a group of hosts, and if for some reason those services are not running, puppet will happily start them up for me. The catch is, in my environment we''re pushing towards IPv6 except for customer facing equipment. So when puppet comes along and restarts a service for me, on the IPv4 interface, everything is kosher. On the IPv6 address, ntpd, rpc.mountd, portmap, et all bind to the same port as my puppet daemon (in this case 18139). It''s not a problem until I go to start puppet, and puppet correctly assumes the port is in use by another running instance until lsof tells me otherwise. Example lsof output: COMMAND PID USER FD TYPE DEVICE SIZE NODE NAME portmap 1598 rpc 9u IPv6 100695714 TCP *:18139 (LISTEN) automount 1672 root 9u IPv6 100695714 TCP *:18139 (LISTEN) rpc.statd 1724 root 9u IPv6 100695714 TCP *:18139 (LISTEN) rpc.idmap 1824 root 9u IPv6 100695714 TCP *:18139 (LISTEN) ruby 32249 root 9u IPv6 100695714 TCP *:18139 (LISTEN) This is under Linux and what from what I''ve read this might not be the case under KAME based IPv6 network stacks (such as FreeBSD). Anybody else seen this? Cheers, Ryan
Ryan, This appears to be the effect of bug #616, see messages from a couple days ago. I think it''s fixed in SVN. Peter Ryan Dooley wrote:> I''m curious if anybody else has seen this behavior.... Right now I have > puppetd making sure that NFS and NTP services are running across a group of > hosts, and if for some reason those services are not running, puppet will > happily start them up for me. > > The catch is, in my environment we''re pushing towards IPv6 except for > customer facing equipment. So when puppet comes along and restarts a > service for me, on the IPv4 interface, everything is kosher. On the IPv6 > address, ntpd, rpc.mountd, portmap, et all bind to the same port as my > puppet daemon (in this case 18139). It''s not a problem until I go to start > puppet, and puppet correctly assumes the port is in use by another running > instance until lsof tells me otherwise. > > Example lsof output: > > COMMAND PID USER FD TYPE DEVICE SIZE NODE NAME > portmap 1598 rpc 9u IPv6 100695714 TCP *:18139 (LISTEN) > automount 1672 root 9u IPv6 100695714 TCP *:18139 (LISTEN) > rpc.statd 1724 root 9u IPv6 100695714 TCP *:18139 (LISTEN) > rpc.idmap 1824 root 9u IPv6 100695714 TCP *:18139 (LISTEN) > ruby 32249 root 9u IPv6 100695714 TCP *:18139 (LISTEN) > > This is under Linux and what from what I''ve read this might not be the case > under KAME based IPv6 network stacks (such as FreeBSD). > > Anybody else seen this? > > Cheers, > Ryan
On Jun 7, 2007, at 11:59 PM, Peter Abrahamsen wrote:> Ryan, > > This appears to be the effect of bug #616, see messages from a couple > days ago. I think it''s fixed in SVN.Indeed it is. -- Honest criticism is hard to take, particularly from a relative, a friend, an acquaintance, or a stranger. -- Franklin P. Jones --------------------------------------------------------------------- Luke Kanies | http://reductivelabs.com | http://madstop.com