[Pulp-list] Repo Sync Scheduling

[Jason L Connor]

On Wed, 2010-06-16 at 11:31 -0400, seth vidal wrote:
> you could have the pulp web site write the info you want out to a db
> then have a process which comes along when the db is changed and
> writes
> out (as root or whatever) the info to /etc/cron.d/
> 
> That way you avoid system calls to crontab which is just pain.
> 
> So the process looks like:
> 
> user logs into webinterface
> user requests some repo be synced
> pulp writes this to a db
> that info is plucked out and written to /etc/cron.d
> 
> that way you're:
> 
> 1. not writing directly to a potentially dangerous path from an apache
> process
> 2. not relying on anything in pulp running - since cron should be
> running all the time
> 3. using the existing system infrastructure 

I think writing into /etc/cron.d is a fine solution, conceptually. But I
thought we were brainstorming on alternatives to this, as apache doesn't
have write permissions to that directory.

If we're going to go this route, it seems to me that the last step has
to be executed by process that has write permissions to that directory.

Personally, I don't really care where this stuff is stored. Only that
tasks gets executed reliably when the customer expects them to.



On another note, I do see a potential caveat with using crontab. We have
to be careful about cleaning up entries if the administrator shuts down
the web services or even removes pulp all together.

-- 
Jason L Connor
Senior Software Engineer
Systems Management and Cloud Enablement
Red Hat, Inc.
+1.919.890.8331
Freenode: linear
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/pulp-list/attachments/20100616/8658afe9/attachment.sig>