[ntp:questions] ntpd and database servers
lsof at nodata.co.uk
Wed Jan 13 19:59:24 UTC 2010
I have had two problems recently that I was unable to solve properly, so
I am turning to the list to help.
I had two database servers where the clock was wrong, by a couple of
hours (too fast). On the first server, ntpd was running but refusing to
sync the clock because of the time difference. On the second server,
ntpd was not running.
These database servers can be stopped quarterly. For data integrity
purposes the clock must not jump backwards. It's more important that the
database server is up and running that if the timestamps are off by a
couple of hours.
Which brings me to my problem: once this problem came to our attention,
we wanted to fix it, slowing the system clock until time caught up. It
wouldn't matter if this took a few weeks.
What is the correct way of solving this? Both on the first server with
ntpd running but out of sync and refusing to update, and on the second
server where ntpd was not running (and I want to avoid a sync on start,
and then avoid the same problem as the first server).
Hope someone can offer their advice for this situation and question.
Note: (sorry to be rude) I'm not looking for an answer which is "plan
downtime", or "except the risk of a clock jumping backwards". Thanks.
More information about the questions