Hello,
I am wondering how I can tell if my PTP infrastructure is healthy or not.
From time to time, our /var/run/ptpd2.status file will show "not
calibrated, in control" for the "Clock status" entry. I assume that's not
good, and "calibrated, in control" is what we want. The
/var/log/ptpd2.stats file shows that Sync messages are coming in regularly
at 1.1-1.6 second intervals. The Delay Response messages are more
irregular, sometimes coming in bunches, and anywhere from 0.006 to 1.9
seconds apart. However, out of 25,045 Delay Response messages, I have noted
7 missing sequence numbers (which may be concerning although doesn't seem
like a lot).
tcpdump/wireshark shows that a lot of my system are sending PTPv2
Management messages to the multicast address 224.0.1.129: "Management Error
Messages (NO_SUCH_ID)", "Management (Clock description) RESPONSE", and
"Management (Current dataset) RESPONSE" are all going to the multicast,
from many different hosts. Is this a problem? I do see "Management (Unknown
management Id 53249) GET" from my Grandmaster clocks, as well as
"Management (Current dataset) GET" multicast messages.
Some Delay Req are going out as multicast from some slave devices. Most of
my PTP slaves are in hybrid mode, so Delay Req are unicast from them. Is it
a problem to mix multicast and unicast Delay Req messages?
Also, I am getting Delay Resp from two clocks on the same network, to the
multicast address. I suppose that's ok? My PTP hybrid slaves that sends
unicast Delay Req are getting unicast Relay_Resp messages from only one of
the clocks, which makes sense to me.
--
-Mike Schwager
Hello,
I am wondering how I can tell if my PTP infrastructure is healthy or not.
From time to time, our /var/run/ptpd2.status file will show "not
calibrated, in control" for the "Clock status" entry. I assume that's not
good, and "calibrated, in control" is what we want. The
/var/log/ptpd2.stats file shows that Sync messages are coming in regularly
at 1.1-1.6 second intervals. The Delay Response messages are more
irregular, sometimes coming in bunches, and anywhere from 0.006 to 1.9
seconds apart. However, out of 25,045 Delay Response messages, I have noted
7 missing sequence numbers (which may be concerning although doesn't seem
like a lot).
tcpdump/wireshark shows that a lot of my system are sending PTPv2
Management messages to the multicast address 224.0.1.129: "Management Error
Messages (NO_SUCH_ID)", "Management (Clock description) RESPONSE", and
"Management (Current dataset) RESPONSE" are all going to the multicast,
from many different hosts. Is this a problem? I do see "Management (Unknown
management Id 53249) GET" from my Grandmaster clocks, as well as
"Management (Current dataset) GET" multicast messages.
Some Delay Req are going out as multicast from some slave devices. Most of
my PTP slaves are in hybrid mode, so Delay Req are unicast from them. Is it
a problem to mix multicast and unicast Delay Req messages?
Also, I am getting Delay Resp from two clocks on the same network, to the
multicast address. I suppose that's ok? My PTP hybrid slaves that sends
unicast Delay Req are getting unicast Relay_Resp messages from only one of
the clocks, which makes sense to me.
--
-Mike Schwager