[Info-vax] Are queue manager updates written to disk immediately ?

[Simon Clubley]

On 2013-04-19, Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> wrote:
>
> I already thought of that one, but thanks anyway for the question.
>
> There was no accounting entry from the first job, but that's not really
> a surprise because accounting entries are either buffered before been
> written to disk or there is a delay in updating the EOF marker.
>
> You can test this for yourself.
>

That came across as a bit abrupt, sorry. (I was heading out the door.). :-)

What I meant to say is that you can test it for yourself in case you
find it hard to believe that I've come across yet another example of
VMS deferring writes to disk. :-)

I've known about accounting logs been deferred for a number of years,
but it's the only part of VMS I _thought_ used deferred updates.

Of course, it's one thing to defer writing away historical data and it's
a completely different situation to defer writing away updates to a active
queue manager database (if that is what is going on).

> In one session, log into a VMS box and have "$ acc/since={10 minutes ago}"
> ready to run.
>
> Now log in (and out again immediately) to a VMS box from another session
> then run the $ acc command above immediately. You will see there is a
> delay before the accounting entry from that session becomes visible.
>

The box I tested it on is a standalone box, not a cluster in case it
makes any difference. I managed to get up to a delay before it became
visible of about 1 minute.

> This is on VMS Alpha v8.3, in case this behaviour is different in other
> VMS versions.
>

Simon.

-- 
Simon Clubley, clubley at remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world