I need to roll out an change of NTP server across approx 200 Linux hosts in a variety of relatively common roles - NGINX. Postfix, MySQL, MongoDB, Redis, RabbitMQ, Greylog for example.
The boxes are broadly using either an upstream internal version using the same clock source, or Debian/Ubuntu pool servers so they should be broadly correct time wise. They are using a combination of ntpd and timesyncd. For the most part I'm not expecting anything too traumatic, so in theory there should be nothing but a small, incremental drift at most following the change, but I'm working if anybody has any best practice advice on that might be helpful in preparing for a smooth migration?
My biggest concern is some clustered or paired services such as MySQL and MongoDB replicated clusters, I recall being told some time ago that they can be a bit more time sensitive. Can anybody tell me if it would be best practice to stop any database or any other services while I make the change to ensure there's no potential impact which the changes are made, or am I being overly cautious?
Thanks!
The boxes are broadly using either an upstream internal version using the same clock source, or Debian/Ubuntu pool servers so they should be broadly correct time wise. They are using a combination of ntpd and timesyncd. For the most part I'm not expecting anything too traumatic, so in theory there should be nothing but a small, incremental drift at most following the change, but I'm working if anybody has any best practice advice on that might be helpful in preparing for a smooth migration?
My biggest concern is some clustered or paired services such as MySQL and MongoDB replicated clusters, I recall being told some time ago that they can be a bit more time sensitive. Can anybody tell me if it would be best practice to stop any database or any other services while I make the change to ensure there's no potential impact which the changes are made, or am I being overly cautious?
Thanks!
