2 questions:
Has anyone set up NTP peering in the service console on ESX?
My ESX servers currently all sync to a symetricom (Stratum 0 GPS NTP appliance) but someone asked me if we could peer them together and I think that the doccumentation says not to do this. Does anyone know why it says not to?
This particular environment runs software that cannot tollerate time differences greater than about 30s and they are worried about what happens if the servers loose contact to the Symetricom, if they were peered they would atleast drift as a unit. Wondering if anyone else has had any experience with this.
--
If you found any of my comments helpful please consider awarding points for "Correct" or "Helpful". Thanks!!!
You can setup several NTP servers, and place your most accurate NTP server first. So if it fails both ESXes will switch to second NTP server.
---
MCSA, MCTS, VCP, VMware vExpert '2009
Not sure exactly what you are considering by 'peer them together'.
As indicated in previous response multiple NTP servers should be configured for redundancy. Consider using ntp.org NTP servers. They have a pool of NTP servers they provide:
The pools of servers are organized by geographic region so you can simply define the following ntp servers for North America for example:
0.north-america.pool.ntp.org
1.north-america.pool.ntp.org
2.north-america.pool.ntp.org
These DNS entries get redirected to various well maintained NTP servers in the regions managed by different organizations and groups so you automatically get redundancy from reliable NTP sources by using the multiple DNS names of the NTP servers in these pools.
I understand how NTP works, I am refering to the "peer" parameter that you can specify in the linux NTP daemon (not ESX specific). Peers are a fallback mechanism in NTP to allow a group of peers to drift together in the event that all outside NTP sources are unreachable. This option does not exist in the GUI but it can be specified directly in the configuration file inside the service console. This is often a practice of large scale *nix networks to make sure that in the event of a total failure all systems drift at the same rate. Hope that makes sense.
Also, this environment does not have access to the outside world at all (physically seperated) so connecting to ntp.org is not an option.
---
If you found any of my comments helpful please consider awarding points for "Correct" or "Helpful". Thanks!!!