thanks mate. Ive done automated installation across the fleet before but it went a lot of technical and process approvals before transition. This migration could encounter the same thing. It seems there is really no other way but do automated restart. Im not sure about running 3 old and 3 new ntp servers in parallel cause I may need to modify clients ntp.conf to add new ntp servers?
On Thu, Jul 7, 2011 at 10:19 PM, andrelst <[email protected]> wrote: > Greek, > I think you are referring to TZ or zoneinfo on Solaris and Linux. And yes, > Solaris needs a reboot since TZ information is stored in memory and Linux, > specifically RHEL, you just run redhat-config-date... no restart needed. > > Philip, > Last time i look at the NTP code, it does a simple call to get the FQDN > first ip address (if there are multiple IP to 1 FQDN) and does not even > bother checking or obey DNS TTL which is I think you are hoping to achieve. > these means no choice but to restart the NTP daemon, which is not a big > deal. > > As you mention 2000 servers, would be very conservative in changing the > aliases on the fly, as variations on each servers for /etc/hosts, > resolv.conf and nsswitch.conf can pretty much guarantee outages. And because > of the variations, it's not certified that new ip address of the aliases > will be picked up. Example, ntp1 has a specific ip address on /etc/hosts. > > Personally, would add server ntp1..6. where you have 3 old NTP and 3 new > NTP running in parallel and do a restart on the NTP daemon. These guarantee > 100% no outage, as you just fix the issues in your leisure time even in > PROD. > > regards, > Andre | http://www.varon.ca > > > On Thu, Jul 7, 2011 at 5:16 AM, Greek Ordono <[email protected]> wrote: > > For Solaris 8/9/10 requires restart and Redhat/Linux reload/SIGHUP > works:P > > -- > Greek Ordono > vmlinuz|genunix|vmkernel admin > myppa: launchpad.net/~grexk/+archive/ppa > From: philip morales <[email protected]> > To: [email protected] > Sent: Thursday, July 7, 2011 4:22:42 PM > Subject: [plug] NTP alias migration > im into migrating very old ntp servers by migrating their aliases into > the new ntp servers. ntp.conf of the clients just points to aliases > let say > server ntp1 > server ntp2 > server ntp3 > Im simulating how long will it take for the clients to pickup the new > ntp servers hostname but my tests using solaris 10 an rhel 5.6. show clients > are still pointing to old ntp servers even if I have migrated aliases but > nslookup is ok on all clients. > > But when I restarted ntpd on the client of course they immediately > showed the correct new ntp servers. > > But its not a good idea to restart ntpd on clients across the fleet. Is > there a way to make clients discover new ntp servers without restarting > their deamon? > > Thanks! > > _________________________________________________ > Philippine Linux Users' Group (PLUG) Mailing List > http://lists.linux.org.ph/mailman/listinfo/plug > Searchable Archives: http://archives.free.net.ph > > _________________________________________________ > Philippine Linux Users' Group (PLUG) Mailing List > http://lists.linux.org.ph/mailman/listinfo/plug > Searchable Archives: http://archives.free.net.ph >
_________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List http://lists.linux.org.ph/mailman/listinfo/plug Searchable Archives: http://archives.free.net.ph
