Clock Drift on rp7410

[Noel Daniel]

Environment - HA Cluster formed by two rp7410s & VA7410.
Problem- There is a gradual drift of time on one of the servers (i.e. lags by twenty minutes for a fortnight with the other server).
Has anyone experience over such an issue & would he/she be good enough to educate me on same.

[Dennis Handly]

You need to use a time server. Searching for that gives a bunch of references:
http://forums1.itrc.hp.com/service/forums/bizsupport/questionanswer.do?threadId=1073839
http://forums1.itrc.hp.com/service/forums/bizsupport/questionanswer.do?threadId=1068248

[Noel Daniel]

See, running a time server to sync the time of the two servers may resolve the issue. But my problem is not that the two servers are not in absolute sync with each other. The issue is, one server tends to lag behind the other gradually, after setting the same time on both servers at the same time. What is the cause for this ? Is there a solution to this ?

[Dennis Handly]

>The issue is, one server tends to lag behind the other gradually, after setting the same time on both servers at the same time. What is the cause for this?

I can only assume it's because computers don't have atomic clocks. :-)

>Is there a solution to this?

That's the main purpose of a time server. It corrects the drift.

[Noel Daniel]

Thanks Dennis for your response.

See as I told you in my first message, the drift is as much as 20 minutes for a fortnight. Can the drift of a "non-atomic" clock be so large : )) ?

[Dennis Handly]

>the drift is as much as 20 minutes for a fortnight. Can the drift of a "non-atomic" clock be so large : )) ?

I've had 20 minutes happen on a g4000. But don't remember how long since I reset it.

[OldSchool]

1) yes, the clocks drift. 2) no they don't all drift the same amount. 3) setting up ntp will correct the problem.

[A. Clay Stephenson]

You need to learn a valuable lesson. Computers are lousy clocks (and thermometers). Just connect to 3 or so internet NTP servers and your boxes easily will maintain time to within a few tens of milliseconds. A few boxes have had firmware patches to correct excessive clock drift so you might check the patch database but really the answer is NTP.

[Carlos M.J.]

Hi Noel:

An NTP client, can be used as an NTP server, the question is if you need to mantain real time or you need to synchronize your servers (for example in a cluster environment).

If you have an NTP server running now, you can use this configuration, but there is a lot of possible configurations on internet.

-Client side-

server 192.168.1.1 version 3 prefer minpoll 3 maxpoll 4


-Server side-

restrict default nomodify notrap
restrict 127.0.0.1
broadcast 192.168.254.255 ttl 4
server
server 127.127.1.0
fudge 127.127.1.0 stratum 3
driftfile /var/lib/ntp/drift
broadcastdelay 0.004
keys /etc/ntp/keys

Hope this helps...

Best regards.

[Carlos M.J.]

Hi Noel:

An NTP client, can be used as an NTP server, the question is if you need to mantain real time or you need to synchronize your servers (for example in a cluster environment).

If you have an NTP server running now, you can use this configuration, but there is a lot of possible configurations on internet.

-Client side-

server 192.168.1.1 version 3 prefer minpoll 3 maxpoll 4


-Server side-

restrict default nomodify notrap
restrict 127.0.0.1
broadcast 192.168.1.255 ttl 4
server
server 127.127.1.0
fudge 127.127.1.0 stratum 3
driftfile /var/lib/ntp/drift
broadcastdelay 0.004
keys /etc/ntp/keys

Hope this helps...

Best regards.

[V. Nyga]

Hi,

maybe a result of a different power supply unit (50 Hz <-> 60 Hz)?

Just a thought
Volkmar

[Romanek]

FYI,

On most HP9000 servers, the L-board contains a battery that is user for multiple things one being the system clock. You may want to schedule HP out to replace it. I have seen this many times as a field engineer.

-Paul