Hi Jake,

Thanks for the suggestion.  I was able to try it out and it seemed to work without an issue for the few times I tried the sequence of events.  My main concern was that my program uses an absolute sleep and if the clock performed a step then it may pause processing for a larger than expected amount of time.   I do have the step adjust enabled at startup though.

I wonder if there is any way to delay the start of ptp4l synchronization just long enough to allow all time information from the master to be received by the NIC?

Thanks,

Jason




On 1/21/2021 5:40 PM, Jacob Keller wrote:

On 1/21/2021 12:53 PM, ja...@astroj.com wrote:
Hello,

I'm currently using v3.1 to sync some Redhat 7.6 machines to a PTP
master.  It has been working well so far but I just noticed a case where
the time will be out of sync until it slewed back slowly after a network
restart only.  When the network goes down by commanding the os to
disable networking, ptp4l and phc2sys recognize this and seem to handle
things properly.  After the network is brought back up, it looks like it
starts to sync to a time without leap seconds just long enough to
generate a large time offset.  ptp4l will detect this and perform a time
step adjustment at this point.  A few moments later the correct time
will be detected on its own and then ptp4l will slew to the correct time
taking a few minutes or more.  Is there a configuration setting that may
handle this disable/enable network jump scenario?

There is a configuration option called step_threshold which defines when
ptp4l will perform a clock step instead of only frequency adjustments.
There is also a similar option that controls behavior only at startup of
the program. If you need to handle steps during when the program runs
you could enable the step_threshold option, as the default setting is to
never step the clock.

Thanks,
Jake



_______________________________________________
Linuxptp-users mailing list
Linuxptp-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-users

Reply via email to