pcp
[Top] [All Lists]

Re: pcp updates - qa and pmlogmv

To: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: pcp updates - qa and pmlogmv
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Thu, 27 Mar 2014 07:00:56 -0400
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <001701cf496f$2a6f52d0$7f4df870$@internode.on.net>
References: <5331FF3C.20304@xxxxxxxxxxxxxxxx> <y0mppl97iff.fsf@xxxxxxxx> <001701cf496f$2a6f52d0$7f4df870$@internode.on.net>
User-agent: Mutt/1.4.2.2i
Hi, Ken -


> [...]  My goal was to ensure we did not clobber or delete parts of
> the archive in the process of renaming, provided the system is still
> running at the end of pmlogmv.

> [...]  The most common use case will be for pmlogger-daily where
> there is already a directory-level lock to prevent bad concurrency
> from the pmlogger-check and pmlogger-daily scripts.  So I don't
> think anything special is needed for this case.

If locking is already coming from another context, then perhaps you
just need to dress up a few plain calls /bin/cp and/or /bin/mv into
pmlogmv.  Perhaps it need only be a subroutine within the pmlogger*
shell scripts.


> For interactive ad hoc use, the risk is no different to multiple
> users/processes executing rm and mv at the same time ... I think
> this is a case of don't point loaded weapons are your feet [...]

That's fine, but use of words like "atomic" can be misinterpreted as
the term of art, as if we were advertising a safe blue training gun
but delivered a real thing.  Maybe replace the word throughout with
"as a group"?


- FChE

<Prev in Thread] Current Thread [Next in Thread>