724642 - admin1b.private.tpe1 can't reach mradm01 internally, causing clock drift

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: NetOps, task)

Description

[Michael Burns [:mburns]]

admin1b.private.tpe1.mozilla.com:Time Sync is WARNING: NTP WARNING: Offset -20.9878484 secs

[[e-mail]1 ~]# grep 10.2.74.123 /etc/ntp.conf 
server 10.2.74.123
[[e-mail]1 ~]# host 10.2.74.123
Host 123.74.2.10.in-addr.arpa. not found: 3(NXDOMAIN)
[[e-mail]1 ~]# host mradm01.mozilla.org
mradm01.mozilla.org has address 63.245.208.161

Comment 1

[Justin Dow [:jabba]]

It shouldn't actually need mradm01. I think I just need to add tpe1 to the facter "datacenter" fact.

Comment 2

[Ravi Pina [:ravi]]

We need to configure the hosts to contact a geographic local NTP server rather than traversing a VPN to a host in a data center in California.

Comment 3

[Justin Dow [:jabba]]

Currently, the ntp puppet module has all remote offices using mvadm01 as their time source. This is mostly important, since they synchronize their LDAP databases from a Mountain View ldap host, also using mvadm01 as its time source. LDAP is the main reason that time sync is critical on these hosts. I'd be ok switching to an external pool as well, but we can configure that as needed in the ntp puppet module, which just uses the "datacenter" facter fact. If we want to change to using local NTP sources in each office, we should open a new bug. This one has been fixed, after I added the "tpe1" fact to puppet.