How to update gradually the system clock via DTSS ?

[darkangel]

Hello

It is me again ;0)

An other seven ES40 nodes is not at a correct time. It is having time shift (1Min 04s late)
for the moment it is under OpenVMS 7.2-1 with TCPIP stack 5.3 eco 2. It is a live server so i can not change abrutely the system time

In few weeks we plan to upgrade the nodes to OpenVMS 7.3-2 with TCPIP stack 5.4 eco 5 on wich NTP will be install.
For the moment i have to reduce the time shift by updating the DTSS process


My problem is i don't understand why i have the following message

%SYSMAN-I-OUTPUT, command execution on node XXXX
%NCL-E-INVALIDABSTIME, invalid absolute time, use \DDDD-HH:MM:SS I SSS\
UPDATE DTSS TIME 2007-02-02:14:53:00

I saw the following topic (the only one)
http://h71000.www7.hp.com/wizard/wiz_3448.html

and i don't understand the help available

NCL>help update dtss

UPDATE

dtss


Gradually adjusts the clock on the specified server node to a new
time, beginning at the time specified by the Time argument.

update [node node-id] dtss time [absolute-time]

Example:

update dtss time 1996-12-30-11:24:00.000-05:00I02.000


What mean the part " -05:00I02.000 " in the help ? how do i apply it to my node ?

I know some command

Using mc ncl synchronize dtss set clock true (will change abrutly the clock) and replacing true by false with change the clock smoothly (what i want)

Thanks for your help.

Br
Darkangel

[Hoff]

Within the DTSS specification of time, the letter I (that's the letter i and not a vertical bar or letter L) introduces the the inaccuracy specification. The value that indicates how far off the time can be; your tolerance for error. For an intro:

http://h71000.www7.hp.com/openvms/journal/v5/decnet.html

For the gonzo description of the construct, look in the DECdts (DTSS) manual:

http://h71000.www7.hp.com/DOC/82final/decnetplus/DTS_MNG.pdf

I might assume you are not utilizing DTSS as a normal matter of course, and that the node involved is not a DTSS server, and are simply seeking to use DTSS to drift the time. If that is the case, I would look to a took such as TBO or SETCLOCK. Both are on the Freeware, though only the latter is listed in the FAQ section with information on how to drift your system time. (I'll add that reference for the next edition of the FAQ.) TBO and SETCLOCK utilize the same OpenVMS mechanisms and interfaces that underlie the DTSS client (DTSS clerk, DECdts clerk, DECdts client -- DTSS calls clients clerks) and the NTP client; these tools drift the clock the same way that DTSS and NTP drift it.

Being somewhat paranoid, I would tend to use and test TBO, SETCLOCK or the DTSS command on a non-production node before I turned any of it loose on a production server.

Stephen Hoffman
HoffmanLabs

[darkangel]

hi

i tried on atest system anymethod to update the node and i don't understand what's happening

For example at 16:40:00 i typed the following command (to have one minute of delay smoothly)

sysman>do mc ncl update dtss time 2007-02-02-16:39:00

and the cluster is having some seconds ahead istead of slowing the clock of one minute.

thanks for your suggestion but in my case i can't do any soft installation . i have to use dtss, nothing less, nothing more

so i thought i can do the same on the production node
it is for example, at 17H20min2s to have one minutes and 2 second ahead, i did

sysman>do mc ncl update dtss time 2007-02-02-17:19:00

and i still have
Node 0 DTSS
at 2007-02-02-17:18:59.720+01:00I91.700
command failed due to:
process failure
Time must be contained in the current time interval.

[Hoff]

Ok, that's a different error message.

Is the production node running a DECdts server?

If it is a DECdts clerk, then it's disconnected from the requisite numbers of DECdts servers, there's a configuration issue, or there's a software bug. (DECdts clerk squawks about insufficient numbers of time servers via OPCOM.) Per the docs, you can't change the time directly on a DECdts clerk, only on a DECdts server.

If this is a DECdts server, specify a new time value that is closer to the current time on the node; you are outside the window permitted by DECdts, and DECdts is declaring your new time value to be unusable per its rules. It's too large a change. Try smaller changes.

Or shut down DECdts clerk and change or drift the time. If it's disconnected from a DECdts server, it's not able to fulfill its basic timekeeping operations.

You could shut down the node and reboot it with the correct time. That'll be faster.

Or load the TBO or SETCLOCK tools and drift it that way. Yes, rules against random production server software installation are good and reasonable, but these rules can and do inevitably have exceptions. Having a seriously bogus system time looks like an exception to me, and a case where installing and auditing a tested and verified repair image is reasonable and acceptable practice. Ask your manager.

Or load and configure and get NTP running and fix this in one shot.

There's another write-up on dtss (DCE DTS, which is very similar to DECdts dtss) here:
http://www.pdc.kth.se/doc/SP/redbooks/pdfbks/sg244616.pdf.gz

Or ring up the HP support folks and ask for help drifting what appears to be a DECdts clerk, or with drifting the time in general.

[darkangel]

Hi

i'll do a smooth manual update by stopping the dtss process then.

Thanks

[darkangel]

does a smooth manual update