Experimental Physics and
Industrial Control System

Maren Purves <[email protected]> · Mon, 27 Nov 2017 16:18:36 -1000

On 11/27/17 16:04, Michael Davidsaver wrote:

On 11/27/2017 02:47 AM, Dirk Zimoch wrote:

On 24.11.2017 17:04, Michael Davidsaver wrote:

On 11/24/2017 01:50 AM, Dirk Zimoch wrote:

On 20.11.2017 11:10, Dirk Zimoch wrote:

Hello everyone,

From time to time we see a strange problem with the timestamp on our
vxWorks 5.5 IOCs. The timestamp jumps to the future (e.g. 2071). I have
seen it happen on EPICS 3.14.12 as well as 3.14.8. Not all but more than
one IOC seem to be affected from such an incident. Not all jump to the
same future time.

Has anyone else seen this effect?
Can a corrupted NTP message cause this?

Dirk

Example:
ZPSAF15-VME > NTPTime_Report 1
NTP driver is synchronized with server
Syncronization interval = 60.0 seconds
Last synchronized at 2017-11-24 08:45:29.149741
OS tick rate = 100 Hz (nominal)
Measured tick rate = 100.024 Hz
NTP Server = 129.129.190.1
value = 0 = 0x0
ZPSAF15-VME > date
2073/05/20 11:20:31.396217
value = 27 = 0x1b

Time jumped this night, only on this IOC.

Are there other time providers in this IOC?  What does
"generalTimeReport 5" show?

The system has been restarted meanwhile, but here is the report:

ZPSAF15-VME > generalTimeReport 5
Backwards time errors prevented 2169 times.

Has anyone else seen this effect?

I saw something like this from an EVR where reading the current time occasionally
resulted in a vme bus error (a firmware issue long fixed).  As a result of this the
mrfioc2 time provider does some fairly strict tests to avoid returning a bad time to
generalTime.  If this sort of thing happens again, it might be worthwhile to add some
additional sanity checking to the NTP provider.

We have seen time going backwards on a Bancom 635 with one time
provider. We have more than one of those boards and only one of
them does that.  The way it is set up (probably in the driver)
is that the year goes ahead (but only by 1 year) when the time
decreases which of course makes telescope pointing go haywire.

If you have several of the same boards/clocks connected the same
way and only one does this I'd blame it on hardware.

Maren

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System